Baseten growth playbook

Rebuild /financial-services/ on the Healthcare page's structure: three to five use cases a bank, insurer or asset manager would recognise (document and research, call transcription and summarisation, fraud and KYC review, coding agents for internal engineering), the compliance and deployment answers a risk team asks first (SOC 2 Type II, self-hosted and region-restricted deployments, data never stored), one proof point, and "Talk to an engineer". Add it to the Industries menu.

An industry page is the first thing a Strategic or Enterprise AE sends after a cold call, and the first thing a bank's architect checks before replying. A stub tells them Baseten has not sold to anyone like them yet.

Title: AI inference for financial services | Baseten
H1: Production AI for banks, insurers and asset managers
1. Use cases: research and document agents; call transcription and summaries; fraud and KYC review; coding agents for internal teams; customer service voice agents
2. What your risk team will ask: where data goes (never stored), where it runs (our cloud, your VPC or hybrid), compliance (SOC 2 Type II), uptime (99.99%), access (RBAC, audit)
3. Proof: one workload, one number
4. Talk to an engineer

• No public financial-services customer yet

  Use an anonymised workload and the compliance answers until a named story is approved.

• Claims a bank's legal team would challenge

  State only what the trust center and contracts support, and link to them.

• Your websiteThe /financial-services/ page is in the sitemap but is a stub: its title is "Financial services", its meta description is "Baseten for financial services", and under the heading "Baseten for financial services" the only call to action reads "Click here", followed by the generic product block. The Healthcare page, by contrast, has use cases, HIPAA and SOC 2 Type II, 99.99% uptime and an OpenEvidence quote. The site's Industries menu lists Enterprise and Healthcare, not Financial Services. The Enterprise page says "Baseten never stores the inputs or outputs of your inference requests".Source
• HiringProduct Marketing Manager, Solutions: "define and own Baseten's up-market, enterprise, and industries go-to-market strategy", build "the vertical go-to-market strategies that let us compete in Financial Services, and Health & Life Sciences"; own industry landing pages, TCO/ROI guides, pricing guides, one-pagers and ABM experiments; "largely greenfield".Source
• Your websitePricing: Basic is "$0 per month, pay as you go"; Pro adds priority GPUs and Slack support; Enterprise adds "Use existing cloud commitments", "Full control over data residency", self-hosting and custom regions. Model APIs are priced per million tokens, for example GLM-5.3 at $1.40 input and $4.40 output and DeepSeek V4.1 Flash at $0.30 and $1.20.Source

Publish /partners with one card per live relationship: what the joint solution is, how a customer buys it (the OpenAI commitment path, cloud marketplace private offers, use of existing cloud commitments from the Enterprise plan), and a contact for partner sellers. Keep the form for new partners underneath.

Partner sellers forward pages, not forms. A page with the buying path on it is the first enablement asset both partner hires will need, and it is the URL to put in every co-marketing piece.

• A partner does not want to be listed yet

  Publish only approved cards and add others as they sign off.

• Your websiteThe footer's "Partners" link goes to /talk-to-us/partner/, a page with two sentences and a form; there is no partner directory, no list of the NVIDIA, cloud marketplace or OpenAI integrations, and /partners/ returns 404.Source
• HiringPartner hires: a Product Marketing Manager, Partnerships to bring "differentiated joint solutions to market" with "NVIDIA, Cloud Hyperscalers, Neocloud providers, and emerging ecosystem players" (solution briefs, partner competitor comparisons, co-marketing calendars), and a Partner Enablement Lead because Baseten "has emerging co-sell relationships with multiple strategic marketplaces and clouds, including OpenAI, NVIDIA and other Hyperscalers, Model Labs, and Data Platforms", to "turn each partnership into a repeatable field motion".Source
• Your websiteOpenAI partnership (September 29, 2026): Baseten is "one of the first open-model inference providers in the OpenAI B2B Marketplace, including a native integration within Codex", and "OpenAI enterprise customers can now use existing OpenAI commitments for open models served by Baseten". The landing page says "Restrictions apply" and asks buyers to join a waitlist; every page carries the banner "Use your OpenAI commitments on Baseten open models".Source
• Your websitePricing: Basic is "$0 per month, pay as you go"; Pro adds priority GPUs and Slack support; Enterprise adds "Use existing cloud commitments", "Full control over data residency", self-hosting and custom regions. Model APIs are priced per million tokens, for example GLM-5.3 at $1.40 input and $4.40 output and DeepSeek V4.1 Flash at $0.30 and $1.20.Source

Replace the link with the clean URL, search the other open posts, docs and email templates for links carrying gclid or utm_medium=paid_search, and add a rule to the new marketing data model that drops paid-search sessions whose referrer is the job board.

The first Marketing Analytics hire is being asked to define how marketing success is measured. Small leaks like this make a paid campaign look better than it is, and they are cheapest to fix before the dashboards exist.

• Tracking that some team relies on

  Swap to owned UTMs rather than removing tracking entirely.

• HiringThe Partner Enablement Lead post links to Baseten's OpenAI page with a full Google Ads click string attached (utm_source=google, utm_medium=paid_search, a campaign ID and a gclid), so visits from the job post can be counted as paid search clicks.Source
• HiringMarketing build-out: Field Marketing Managers in San Francisco and New York ("prioritize seller requests by expected revenue"); a Social Media Manager because "most of them meet us first on X or LinkedIn, around a model launch, a benchmark, an open source release, or a customer result" and "we post about latency, throughput, and GPU cost, and we hold ourselves to getting those details right"; the first Marketing Analytics Manager because "Marketing at Baseten is scaling fast: more spend, more campaigns, more model launches and more inbound"; a Marketing Technology & Operations Lead and a GTM Engineer.Source

For each target account, the SDR pre-fills an estimate from what is public (the closed model the product uses, rough request volume from the AE's research, the workload type) and sends one number with a link to rerun it with their own CSV. The reply ask is a 20-minute review with a Solutions Architect.

A cold email with a dollar figure about the prospect's own workload earns replies that a generic intro does not, and the calculator's CSV upload turns the reply into a qualified conversation.

• Estimates off by a lot

  Present a range and say what it assumes; the CSV rerun corrects it.

• Feels presumptuous

  Frame it as "what teams like yours usually see", not a claim about their bill.

• Your websiteThe Open-Source ROI Calculator compares spend on closed models (GPT 5.6, Claude Opus 5, Gemini 3.x and others) with open models on Baseten Model APIs, takes monthly spend or an uploaded usage CSV, and asks the workload type (chat, document processing, agentic). It carries a disclaimer that results are estimates.Source
• HiringSales is split into segments: AI Native AEs (San Francisco, New York) and a Sales Manager for "our fastest growing customer segment"; Enterprise AEs in San Francisco, New York and London; Strategic AEs and Strategic Sales Managers; inbound and outbound SDRs (outbound works "a defined set of target accounts" with an AE by cold call, email and LinkedIn) and a Sales Development Manager. The outbound SDR post has been open since February 2025.Source
• Your websitePricing: Basic is "$0 per month, pay as you go"; Pro adds priority GPUs and Slack support; Enterprise adds "Use existing cloud commitments", "Full control over data residency", self-hosting and custom regions. Model APIs are priced per million tokens, for example GLM-5.3 at $1.40 input and $4.40 output and DeepSeek V4.1 Flash at $0.30 and $1.20.Source

A fixed one-page template sent to every AE, SA and SDR within 24 hours of a launch: what shipped, the one performance claim and how it was measured, which customer types to call this week, a three-line email, and the objection to expect. The same page feeds the social post.

Sellers who can call the right accounts on launch day turn attention into meetings; a template means the brief takes an hour to fill in rather than a week to design.

• Brief goes stale fast

  Date it and replace it at the next launch rather than editing.

• HiringLaunch roles: the first Product Enablement Lead, to own "how these launches land with the field" for market moments such as the GLM-5.2 and Kimi K3 launches; a Product Marketing Manager, Post-Training to "shape how the market thinks about closed-to-open weights models and post-training ROI"; and a Product Marketing Manager, Dedicated Inference for autoscaling, routing, failover and compliance stories.Source
• HiringMarketing build-out: Field Marketing Managers in San Francisco and New York ("prioritize seller requests by expected revenue"); a Social Media Manager because "most of them meet us first on X or LinkedIn, around a model launch, a benchmark, an open source release, or a customer result" and "we post about latency, throughput, and GPU cost, and we hold ourselves to getting those details right"; the first Marketing Analytics Manager because "Marketing at Baseten is scaling fast: more spend, more campaigns, more model launches and more inbound"; a Marketing Technology & Operations Lead and a GTM Engineer.Source
• Your websiteHomepage: two H1 headings ("Inference is everything" and "The fastest inference takes more than GPUs."); logos include Abridge, Canva, Cursor, Datadog, Decagon, Harvey, HubSpot, Lovable, Notion, OpenEvidence, Parallel, Ramp, Vercel and Wayfair; claims such as "Baseten Embeddings Inference (BEI) has over 2x higher throughput and 10% lower latency than any other solution on the market" appear without a linked method.Source

Create one public methods page: for each performance claim on the site, the model, hardware, runtime, batch size, date and the comparison set, plus a way to reproduce it. Link every claim to its entry, and require an entry before a claim goes on a page or a post.

For a technical audience, a claim with a method is persuasive and a claim without one invites a rebuttal thread. It also gives sellers a link to send when a prospect pushes back.

• Publishing methods helps competitors

  Methods are already inferable; credibility is worth more than secrecy here.

• A claim turns out to be stale

  Update or remove it; that is the point of the page.

• Your websiteHomepage: two H1 headings ("Inference is everything" and "The fastest inference takes more than GPUs."); logos include Abridge, Canva, Cursor, Datadog, Decagon, Harvey, HubSpot, Lovable, Notion, OpenEvidence, Parallel, Ramp, Vercel and Wayfair; claims such as "Baseten Embeddings Inference (BEI) has over 2x higher throughput and 10% lower latency than any other solution on the market" appear without a linked method.Source
• HiringMarketing build-out: Field Marketing Managers in San Francisco and New York ("prioritize seller requests by expected revenue"); a Social Media Manager because "most of them meet us first on X or LinkedIn, around a model launch, a benchmark, an open source release, or a customer result" and "we post about latency, throughput, and GPU cost, and we hold ourselves to getting those details right"; the first Marketing Analytics Manager because "Marketing at Baseten is scaling fast: more spend, more campaigns, more model launches and more inbound"; a Marketing Technology & Operations Lead and a GTM Engineer.Source
• Competitor researchRivals' own positioning: Together AI calls itself "The AI Native Cloud"; Fireworks' homepage title is "Own Your Specialized Intelligence" with a "training and inference platform"; Modal is "The platform for production AI"; Replicate offers "Run AI with an API". Baseten names its own fastest-growing sales segment "AI Native" and its Series F post speaks of customers who want to "own their intelligence".See the research Source

Before the sellers start: an EMEA page (EU and UK regions, data residency through region-restricted deployments, GDPR and the DPA, euro and sterling contracting, the London team), three European customer stories from the accounts already served from the US, and a London launch event co-hosted with one of them.

European buyers ask where data runs and who else in Europe uses the vendor before anything else. Twenty sellers without those answers will spend their first quarter writing them one email at a time.

• Customers decline to be named

  Publish anonymised stories with region and workload.

• Residency promises outrun the product

  State only regions and options available today.

• HiringHead of EMEA Sales (London, posted September 14, 2026): a "Founding Sales Leader in EMEA to build our business in the region from the ground up". "We already have EMEA customers and real demand — today it's served from the US." The mandate is "to scale the EMEA team to 20+ people this year", carry the number while hiring, and "adapt the motion" to how EMEA technical buyers evaluate infrastructure. London is also hiring a Strategic AE, an Enterprise AE, a Strategic Sales Manager and an Enterprise Sales Manager, plus a remote EMEA Solutions Architect.Source
• Your websiteCustomer stories: 27 in the sitemap, mostly AI-native companies (Notion, Gamma, Superhuman, Writer, Zed, OpenCode, Wispr Flow, Rime, Speechify) and healthcare (OpenEvidence twice, Sully.ai, Latent); none from a bank, insurer or asset manager, and none from a European company. The Notion story's address reads "how-notion-acheived-2-3x-lower-latency".Source
• Your websiteSitemap (1,236 URLs): /resources 405, /library 357 model pages, /blog 254, /author 117 author pages and a 49-page "Inference Engineering" book. The only Europe-specific content is a post on running DeepSeek R1 "in US & EU data centers"; there is no UK or EU page. Event pages include AWS re:Invent parties co-hosted with Notion, LangChain and Duckbill.Source
• Your websitePricing: Basic is "$0 per month, pay as you go"; Pro adds priority GPUs and Slack support; Enterprise adds "Use existing cloud commitments", "Full control over data residency", self-hosting and custom regions. Model APIs are priced per million tokens, for example GLM-5.3 at $1.40 input and $4.40 output and DeepSeek V4.1 Flash at $0.30 and $1.20.Source

Treat every waitlist entry as an opportunity: qualify within two days (commitment size, Codex usage, which open model they want), give an SA-led trial with a cost comparison against the closed model they use, and build a joint account list with OpenAI's field for the largest Codex customers.

The commitment removes the procurement step, which is usually the slowest part of an enterprise infrastructure deal. The motion only works if Baseten moves faster than the buyer's interest fades.

• OpenAI restrictions limit who qualifies

  Qualify against the actual rules first and tell non-qualifying accounts about direct pricing.

• Your websiteOpenAI partnership (September 29, 2026): Baseten is "one of the first open-model inference providers in the OpenAI B2B Marketplace, including a native integration within Codex", and "OpenAI enterprise customers can now use existing OpenAI commitments for open models served by Baseten". The landing page says "Restrictions apply" and asks buyers to join a waitlist; every page carries the banner "Use your OpenAI commitments on Baseten open models".Source
• HiringPartner hires: a Product Marketing Manager, Partnerships to bring "differentiated joint solutions to market" with "NVIDIA, Cloud Hyperscalers, Neocloud providers, and emerging ecosystem players" (solution briefs, partner competitor comparisons, co-marketing calendars), and a Partner Enablement Lead because Baseten "has emerging co-sell relationships with multiple strategic marketplaces and clouds, including OpenAI, NVIDIA and other Hyperscalers, Model Labs, and Data Platforms", to "turn each partnership into a repeatable field motion".Source
• Your websiteThe Open-Source ROI Calculator compares spend on closed models (GPT 5.6, Claude Opus 5, Gemini 3.x and others) with open models on Baseten Model APIs, takes monthly spend or an uploaded usage CSV, and asks the workload type (chat, document processing, agentic). It carries a disclaimer that results are estimates.Source

Three named stories that each show a customer moving a task from a closed model to a post-trained open one, with quality on their eval, cost per task and latency before and after; one aggregate page stating the post-trained share of traffic; and a short guide on when post-training pays off and when it does not.

Buyers considering open models fear losing quality. Customer evidence plus the platform-wide share makes "post-train and own it" feel like the common path rather than a bet.

• Customers treat their eval results as secret

  Publish relative improvements and the method, not raw scores.

• HiringPost-Training Research Engineer: "By inference volume, the overwhelming majority of traffic at Baseten is to and from models that have been post-trained in some way". The Applied Researcher post: "post-train open-source models that outperform frontier closed models on their specialised tasks" in "healthcare, code generation, and legal".Source
• HiringLaunch roles: the first Product Enablement Lead, to own "how these launches land with the field" for market moments such as the GLM-5.2 and Kimi K3 launches; a Product Marketing Manager, Post-Training to "shape how the market thinks about closed-to-open weights models and post-training ROI"; and a Product Marketing Manager, Dedicated Inference for autoscaling, routing, failover and compliance stories.Source
• Your websiteCustomer stories: 27 in the sitemap, mostly AI-native companies (Notion, Gamma, Superhuman, Writer, Zed, OpenCode, Wispr Flow, Rime, Speechify) and healthcare (OpenEvidence twice, Sully.ai, Latent); none from a bank, insurer or asset manager, and none from a European company. The Notion story's address reads "how-notion-acheived-2-3x-lower-latency".Source
• Your websiteSeries F post (June 22, 2026): "$1.5B Series F" at "a $13B valuation", led by Altimeter Capital, Conviction Partners and Spark Capital; "Over the last year, our revenue has grown 20x, and inference volume has grown 40x"; customers "more motivated than ever to own their intelligence".Source

Run a message test with ten customers and ten prospects: which phrases they associate with which vendor, and which proof makes Baseten different (production inference performance, forward-deployed engineers, multi-cloud capacity). Pick one line Baseten can own and prove, and use it consistently on the homepage, in launches and in the Series F follow-ups.

When two vendors use the same words, buyers fall back to price and brand recognition. A distinct claim backed by proof is what the new PMMs and the field can repeat.

• Rewording after a funding announcement looks unsteady

  Keep the idea, sharpen the words; no rebrand needed.

• Competitor researchRivals' own positioning: Together AI calls itself "The AI Native Cloud"; Fireworks' homepage title is "Own Your Specialized Intelligence" with a "training and inference platform"; Modal is "The platform for production AI"; Replicate offers "Run AI with an API". Baseten names its own fastest-growing sales segment "AI Native" and its Series F post speaks of customers who want to "own their intelligence".See the research Source
• Your websiteSeries F post (June 22, 2026): "$1.5B Series F" at "a $13B valuation", led by Altimeter Capital, Conviction Partners and Spark Capital; "Over the last year, our revenue has grown 20x, and inference volume has grown 40x"; customers "more motivated than ever to own their intelligence".Source
• HiringLaunch roles: the first Product Enablement Lead, to own "how these launches land with the field" for market moments such as the GLM-5.2 and Kimi K3 launches; a Product Marketing Manager, Post-Training to "shape how the market thinks about closed-to-open weights models and post-training ROI"; and a Product Marketing Manager, Dedicated Inference for autoscaling, routing, failover and compliance stories.Source

Define three usage triggers (sustained spend above a threshold, a second model in production, latency or rate-limit errors near plan limits), send each to the account's AE with context, and give the AE a standard offer: a dedicated deployment review with an SA.

Usage-based accounts outgrow shared APIs in a predictable way. Calling at that moment, with the customer's own numbers, converts better than outreach on a schedule.

• Too many alerts

  Start with the top trigger only and raise thresholds until AEs act on every alert.

• HiringSales is split into segments: AI Native AEs (San Francisco, New York) and a Sales Manager for "our fastest growing customer segment"; Enterprise AEs in San Francisco, New York and London; Strategic AEs and Strategic Sales Managers; inbound and outbound SDRs (outbound works "a defined set of target accounts" with an AE by cold call, email and LinkedIn) and a Sales Development Manager. The outbound SDR post has been open since February 2025.Source
• HiringMarketing build-out: Field Marketing Managers in San Francisco and New York ("prioritize seller requests by expected revenue"); a Social Media Manager because "most of them meet us first on X or LinkedIn, around a model launch, a benchmark, an open source release, or a customer result" and "we post about latency, throughput, and GPU cost, and we hold ourselves to getting those details right"; the first Marketing Analytics Manager because "Marketing at Baseten is scaling fast: more spend, more campaigns, more model launches and more inbound"; a Marketing Technology & Operations Lead and a GTM Engineer.Source
• Your websitePricing: Basic is "$0 per month, pay as you go"; Pro adds priority GPUs and Slack support; Enterprise adds "Use existing cloud commitments", "Full control over data residency", self-hosting and custom regions. Model APIs are priced per million tokens, for example GLM-5.3 at $1.40 input and $4.40 output and DeepSeek V4.1 Flash at $0.30 and $1.20.Source

A monthly evening in each city co-hosted with one customer: a 20-minute technical talk from the customer's engineers, an SA on the same topic, 40 invited ML leads from target accounts. Each event gets a target-account list from the AEs before invites go out and a follow-up within two days.

Engineers come to hear peers, not vendors. A customer co-host makes the invite credible and gives the AE a reference without asking for one.

• Same faces every month

  Rotate co-hosts and topics; cap repeat attendees.

• Your websiteSitemap (1,236 URLs): /resources 405, /library 357 model pages, /blog 254, /author 117 author pages and a 49-page "Inference Engineering" book. The only Europe-specific content is a post on running DeepSeek R1 "in US & EU data centers"; there is no UK or EU page. Event pages include AWS re:Invent parties co-hosted with Notion, LangChain and Duckbill.Source
• HiringMarketing build-out: Field Marketing Managers in San Francisco and New York ("prioritize seller requests by expected revenue"); a Social Media Manager because "most of them meet us first on X or LinkedIn, around a model launch, a benchmark, an open source release, or a customer result" and "we post about latency, throughput, and GPU cost, and we hold ourselves to getting those details right"; the first Marketing Analytics Manager because "Marketing at Baseten is scaling fast: more spend, more campaigns, more model launches and more inbound"; a Marketing Technology & Operations Lead and a GTM Engineer.Source
• Your websiteHomepage: two H1 headings ("Inference is everything" and "The fastest inference takes more than GPUs."); logos include Abridge, Canva, Cursor, Datadog, Decagon, Harvey, HubSpot, Lovable, Notion, OpenEvidence, Parallel, Ramp, Vercel and Wayfair; claims such as "Baseten Embeddings Inference (BEI) has over 2x higher throughput and 10% lower latency than any other solution on the market" appear without a linked method.Source

Add a life sciences section (research search, document review, trial operations) built on the Latent story, and a downloadable buyer's guide: how HIPAA and data handling work on Baseten, deployment options, the questions procurement asks and how a pilot is scoped.

Regulated buyers move when the review board has its answers in one document. The guide doubles as the TCO and compliance asset the PMM is asked to build.

• Overstating compliance

  Security signs off every line.

• HiringProduct Marketing Manager, Solutions: "define and own Baseten's up-market, enterprise, and industries go-to-market strategy", build "the vertical go-to-market strategies that let us compete in Financial Services, and Health & Life Sciences"; own industry landing pages, TCO/ROI guides, pricing guides, one-pagers and ABM experiments; "largely greenfield".Source
• Your websiteThe /financial-services/ page is in the sitemap but is a stub: its title is "Financial services", its meta description is "Baseten for financial services", and under the heading "Baseten for financial services" the only call to action reads "Click here", followed by the generic product block. The Healthcare page, by contrast, has use cases, HIPAA and SOC 2 Type II, 99.99% uptime and an OpenEvidence quote. The site's Industries menu lists Enterprise and Healthcare, not Financial Services. The Enterprise page says "Baseten never stores the inputs or outputs of your inference requests".Source
• Your websiteCustomer stories: 27 in the sitemap, mostly AI-native companies (Notion, Gamma, Superhuman, Writer, Zed, OpenCode, Wispr Flow, Rime, Speechify) and healthcare (OpenEvidence twice, Sully.ai, Latent); none from a bank, insurer or asset manager, and none from a European company. The Notion story's address reads "how-notion-acheived-2-3x-lower-latency".Source

Give every library page a standard top block: per-million-token price, measured speed with a link to the methods page, a one-click Try it, a code snippet, and a comparison with the closed model it usually replaces. Publish new model pages on release day and submit them for indexing.

Release-day searches are high-intent and short-lived. A complete page that loads first collects sign-ups that would otherwise go to whichever provider ranks.

• Thin, near-duplicate pages

  Each page carries real price, speed and code, not just a name.

• Your websiteSitemap (1,236 URLs): /resources 405, /library 357 model pages, /blog 254, /author 117 author pages and a 49-page "Inference Engineering" book. The only Europe-specific content is a post on running DeepSeek R1 "in US & EU data centers"; there is no UK or EU page. Event pages include AWS re:Invent parties co-hosted with Notion, LangChain and Duckbill.Source
• Your websitePricing: Basic is "$0 per month, pay as you go"; Pro adds priority GPUs and Slack support; Enterprise adds "Use existing cloud commitments", "Full control over data residency", self-hosting and custom regions. Model APIs are priced per million tokens, for example GLM-5.3 at $1.40 input and $4.40 output and DeepSeek V4.1 Flash at $0.30 and $1.20.Source
• HiringMarketing build-out: Field Marketing Managers in San Francisco and New York ("prioritize seller requests by expected revenue"); a Social Media Manager because "most of them meet us first on X or LinkedIn, around a model launch, a benchmark, an open source release, or a customer result" and "we post about latency, throughput, and GPU cost, and we hold ourselves to getting those details right"; the first Marketing Analytics Manager because "Marketing at Baseten is scaling fast: more spend, more campaigns, more model launches and more inbound"; a Marketing Technology & Operations Lead and a GTM Engineer.Source

Flag every startup-program account 60 days before credits expire or when monthly usage crosses a threshold, and have an AI Native AE offer a short architecture review and a committed-use price. Track which cohort converts.

The credit end date is a known decision point. Meeting it with a person and a price keeps the startups that grew on Baseten from re-shopping at the moment they are largest.

• Startups feel sold to

  Lead with the architecture review; price comes second.

• HiringSales is split into segments: AI Native AEs (San Francisco, New York) and a Sales Manager for "our fastest growing customer segment"; Enterprise AEs in San Francisco, New York and London; Strategic AEs and Strategic Sales Managers; inbound and outbound SDRs (outbound works "a defined set of target accounts" with an AE by cold call, email and LinkedIn) and a Sales Development Manager. The outbound SDR post has been open since February 2025.Source
• Your websitePricing: Basic is "$0 per month, pay as you go"; Pro adds priority GPUs and Slack support; Enterprise adds "Use existing cloud commitments", "Full control over data residency", self-hosting and custom regions. Model APIs are priced per million tokens, for example GLM-5.3 at $1.40 input and $4.40 output and DeepSeek V4.1 Flash at $0.30 and $1.20.Source
• Your websiteSeries F post (June 22, 2026): "$1.5B Series F" at "a $13B valuation", led by Altimeter Capital, Conviction Partners and Spark Capital; "Over the last year, our revenue has grown 20x, and inference volume has grown 40x"; customers "more motivated than ever to own their intelligence".Source

A free self-paced course built on the book with hands-on labs on Baseten, a short certification exam, and a version run live in New York, San Francisco and London. Certified engineers get a public badge and an invitation to a practitioners' group.

Teams adopt the tools they learned on. A certification spreads that learning inside target accounts and gives the field a reason to contact engineers who finish it.

• A course costs more to keep current than a book

  Version it with the book and update labs at each major release.

• Your websiteSitemap (1,236 URLs): /resources 405, /library 357 model pages, /blog 254, /author 117 author pages and a 49-page "Inference Engineering" book. The only Europe-specific content is a post on running DeepSeek R1 "in US & EU data centers"; there is no UK or EU page. Event pages include AWS re:Invent parties co-hosted with Notion, LangChain and Duckbill.Source
• HiringMarketing build-out: Field Marketing Managers in San Francisco and New York ("prioritize seller requests by expected revenue"); a Social Media Manager because "most of them meet us first on X or LinkedIn, around a model launch, a benchmark, an open source release, or a customer result" and "we post about latency, throughput, and GPU cost, and we hold ourselves to getting those details right"; the first Marketing Analytics Manager because "Marketing at Baseten is scaling fast: more spend, more campaigns, more model launches and more inbound"; a Marketing Technology & Operations Lead and a GTM Engineer.Source
• HiringHead of EMEA Sales (London, posted September 14, 2026): a "Founding Sales Leader in EMEA to build our business in the region from the ground up". "We already have EMEA customers and real demand — today it's served from the US." The mandate is "to scale the EMEA team to 20+ people this year", carry the number while hiring, and "adapt the motion" to how EMEA technical buyers evaluate infrastructure. London is also hiring a Strategic AE, an Enterprise AE, a Strategic Sales Manager and an Enterprise Sales Manager, plus a remote EMEA Solutions Architect.Source

Sell capacity as an outcome: a reserved-capacity offer with guaranteed GPU types and regions across Baseten's cloud pool, priced on commitment, sold by Strategic AEs with Supply at the table, and positioned for enterprises whose own cloud allocation is the bottleneck.

When capacity is scarce, the vendor that can guarantee it wins the deal before performance is compared. Baseten's supply work becomes a sales advantage instead of a cost centre.

• Committing to capacity Baseten cannot fill

  Match each customer commitment to a supply contract before signing.

• HiringA Supply team buying compute: Capacity Managers, a Data Centers & Powered Land Sourcing Lead, an NVIDIA, OEM & Chip Providers Lead, a Hyperscalers & Strategic Neoclouds Lead and an Emerging Clouds & International Coverage Lead ("who's building, who's buying, and on what terms").Source
• Your websitePricing: Basic is "$0 per month, pay as you go"; Pro adds priority GPUs and Slack support; Enterprise adds "Use existing cloud commitments", "Full control over data residency", self-hosting and custom regions. Model APIs are priced per million tokens, for example GLM-5.3 at $1.40 input and $4.40 output and DeepSeek V4.1 Flash at $0.30 and $1.20.Source
• HiringSales is split into segments: AI Native AEs (San Francisco, New York) and a Sales Manager for "our fastest growing customer segment"; Enterprise AEs in San Francisco, New York and London; Strategic AEs and Strategic Sales Managers; inbound and outbound SDRs (outbound works "a defined set of target accounts" with an AE by cold call, email and LinkedIn) and a Sales Development Manager. The outbound SDR post has been open since February 2025.Source

An annual report with anonymised aggregates: which model families are served, the share of post-trained traffic, latency and cost trends by modality, and how fast new open models are adopted after release. Release it with customer commentary and a live session.

Original data earns press, analyst citations and links that product pages never get, and it frames Baseten as the place where the open-model market is visible.

• Data gives rivals insight

  Publish trends and shares, not absolute volumes.

• Your websiteSeries F post (June 22, 2026): "$1.5B Series F" at "a $13B valuation", led by Altimeter Capital, Conviction Partners and Spark Capital; "Over the last year, our revenue has grown 20x, and inference volume has grown 40x"; customers "more motivated than ever to own their intelligence".Source
• HiringPost-Training Research Engineer: "By inference volume, the overwhelming majority of traffic at Baseten is to and from models that have been post-trained in some way". The Applied Researcher post: "post-train open-source models that outperform frontier closed models on their specialised tasks" in "healthcare, code generation, and legal".Source
• Your websiteSitemap (1,236 URLs): /resources 405, /library 357 model pages, /blog 254, /author 117 author pages and a 49-page "Inference Engineering" book. The only Europe-specific content is a post on running DeepSeek R1 "in US & EU data centers"; there is no UK or EU page. Event pages include AWS re:Invent parties co-hosted with Notion, LangChain and Duckbill.Source

A fixed offer: in 48 hours a Baseten FDE and the customer's engineers move one production task from a closed model to an open one, with an eval harness comparing quality, cost and latency. If the result is not better on two of three, the customer walks away with the harness.

A time-boxed, low-risk trial turns "someday" into a decision this week, and the eval harness is useful to the customer either way.

• FDE time is scarce

  Qualify sprints by account size and cap them monthly.

• Your websiteThe Open-Source ROI Calculator compares spend on closed models (GPT 5.6, Claude Opus 5, Gemini 3.x and others) with open models on Baseten Model APIs, takes monthly spend or an uploaded usage CSV, and asks the workload type (chat, document processing, agentic). It carries a disclaimer that results are estimates.Source
• HiringPost-Training Research Engineer: "By inference volume, the overwhelming majority of traffic at Baseten is to and from models that have been post-trained in some way". The Applied Researcher post: "post-train open-source models that outperform frontier closed models on their specialised tasks" in "healthcare, code generation, and legal".Source
• Your websiteOpenAI partnership (September 29, 2026): Baseten is "one of the first open-model inference providers in the OpenAI B2B Marketplace, including a native integration within Codex", and "OpenAI enterprise customers can now use existing OpenAI commitments for open models served by Baseten". The landing page says "Restrictions apply" and asks buyers to join a waitlist; every page carries the banner "Use your OpenAI commitments on Baseten open models".Source

Let any developer follow a model family on its library page; when Baseten serves a new version, send one email with price, speed, the methods link and a one-click Try it. The followers of each model are also a ready-made list for the field brief.

An email to someone who asked for it converts far better than a post in a feed, and the follow list shows sellers which accounts are watching which models.

• Too many emails

  One email per family release, no digests.

• HiringMarketing build-out: Field Marketing Managers in San Francisco and New York ("prioritize seller requests by expected revenue"); a Social Media Manager because "most of them meet us first on X or LinkedIn, around a model launch, a benchmark, an open source release, or a customer result" and "we post about latency, throughput, and GPU cost, and we hold ourselves to getting those details right"; the first Marketing Analytics Manager because "Marketing at Baseten is scaling fast: more spend, more campaigns, more model launches and more inbound"; a Marketing Technology & Operations Lead and a GTM Engineer.Source
• Your websiteSitemap (1,236 URLs): /resources 405, /library 357 model pages, /blog 254, /author 117 author pages and a 49-page "Inference Engineering" book. The only Europe-specific content is a post on running DeepSeek R1 "in US & EU data centers"; there is no UK or EU page. Event pages include AWS re:Invent parties co-hosted with Notion, LangChain and Duckbill.Source
• HiringLaunch roles: the first Product Enablement Lead, to own "how these launches land with the field" for market moments such as the GLM-5.2 and Kimi K3 launches; a Product Marketing Manager, Post-Training to "shape how the market thinks about closed-to-open weights models and post-training ROI"; and a Product Marketing Manager, Dedicated Inference for autoscaling, routing, failover and compliance stories.Source

The GTM Engineer adds a job-post signal to the enrichment pipeline: target accounts posting inference, ML platform or post-training roles get a score bump and a same-week SDR email that offers a free architecture review and the Inference Engineering book.

A hiring post is a public, dated signal of a budgeted project. Reaching the team while they are still deciding the stack is the cheapest moment to win.

• Recruiters, not buyers, reply

  Address the email to the hiring manager named in the post or the ML lead.

• HiringSales is split into segments: AI Native AEs (San Francisco, New York) and a Sales Manager for "our fastest growing customer segment"; Enterprise AEs in San Francisco, New York and London; Strategic AEs and Strategic Sales Managers; inbound and outbound SDRs (outbound works "a defined set of target accounts" with an AE by cold call, email and LinkedIn) and a Sales Development Manager. The outbound SDR post has been open since February 2025.Source
• HiringMarketing build-out: Field Marketing Managers in San Francisco and New York ("prioritize seller requests by expected revenue"); a Social Media Manager because "most of them meet us first on X or LinkedIn, around a model launch, a benchmark, an open source release, or a customer result" and "we post about latency, throughput, and GPU cost, and we hold ourselves to getting those details right"; the first Marketing Analytics Manager because "Marketing at Baseten is scaling fast: more spend, more campaigns, more model launches and more inbound"; a Marketing Technology & Operations Lead and a GTM Engineer.Source
• Your websiteSitemap (1,236 URLs): /resources 405, /library 357 model pages, /blog 254, /author 117 author pages and a 49-page "Inference Engineering" book. The only Europe-specific content is a post on running DeepSeek R1 "in US & EU data centers"; there is no UK or EU page. Event pages include AWS re:Invent parties co-hosted with Notion, LangChain and Duckbill.Source