Baseten Just Named Its Recruiting Shortlist. 25 Logos. 7,600 Engineers.
Baseten's $1.5B Series F triples headcount. The customer roster (Cursor, Notion, Abridge, Harvey) is the sourcing map every inference platform now works.
Baseten closed a $1.5B Series F on June 22, 2026 at valuations up to $13B, and used the announcement to promise a tripling of headcount across engineering, research, ops, and GTM. The press cycle focused on the check size. The more useful artifact for anyone recruiting in AI infrastructure is the customer roster Baseten shipped alongside it: Cursor, Notion, Abridge, Harvey, Clay, Lovable, Decagon, Writer, HubSpot, OpenEvidence, Parallel, Gamma, plus another dozen names. That list is a pre-vetted shortlist of every company already running production inference at scale, and it is equally usable by every Baseten competitor now hiring against them.
The round is a hiring round, and the map came with it
The Series F was led by Altimeter, Conviction, and Spark, with Sands Capital and Wellington co-leading. Baseten stated plainly that the money is going into engineering, research, operations, and GTM. The careers page currently lists roughly 63 open roles across four departments: 34 in EPD, 13 in G&A, 13 in GTM, and 3 in talent. Named openings include GPU Kernel Engineer, Machine Learning Engineer for Fine Tuning, Forward Deployed Engineer for EMEA, AI Support Engineer, and a software engineering intern slot in SF and NYC.
Tripling from Baseten's current base is a large ask in any market. It becomes surreal when you look at the pool. Refolk's index shows roughly 7,600 profiles worldwide who combine CUDA, PyTorch, or Triton with inference, ML-infra, GPU, or perf-engineer titles. That is the total addressable universe. Baseten is not the only buyer.
Who else is reading the same PR
Modal, Fireworks, Together AI, Anyscale, Replicate, RunPod, and the hyperscaler inference orgs (Bedrock, Vertex, Azure AI Foundry) are working the same list this week. Fireworks in particular is a mirror image of Baseten's open-source-inference thesis, and Sacra estimates it hit roughly $800M annualized revenue in May 2026, ahead of Baseten's $600M. Fireworks last raised at a $552M valuation in late 2024, which matters for the equity conversation below. The point is that the Baseten Series F hiring surge is not happening in a quiet market. Every well-funded platform is looking at the same customer logos.
Why the customer list beats the org chart
The naive move is to source directly from Baseten's competitors. That misses where the actual work is being done. Baseten's best future hires are sitting inside Baseten's customers, where they have spent the last 12 to 24 months hitting production limits at Cursor, Notion, and Abridge scale. Those engineers have already solved the problems Baseten's own platform is trying to abstract.
Consider the specific pockets:
Cursor routes inference through Baseten for parts of its coding assistant and crossed roughly $2B ARR earlier this year. A senior IC on Cursor's inference team has shipped low-latency code completion under real user load. That resume is worth three hires from a generic ML platform team.
OpenEvidence runs billions of fine-tuned LLM calls per week for clinicians in every major U.S. healthcare facility. Engineers there have solved HIPAA constraints, latency, and fine-tuning simultaneously. There is no cleaner pool for regulated-inference talent.
Abridge, Ambience, and PicnicHealth form a clinical-AI cluster with overlapping technical requirements. If you are staffing a healthcare-facing inference platform, that is three companies of ideal candidates.
Harvey and Hebbia cover legal and finance AI. Engineers in these orgs have publicly shipped roughly 2.5x throughput improvements and 10x cost reductions versus closed-model providers. That is exactly the operator profile inference platforms want.
Notion, HubSpot, Writer, Gamma, Lovable, Decagon, Clay, Parallel round out the app-layer set. Every one of them is running post-trained or custom models in production, which is where Baseten CEO Tuhin Srivastava said "post-training has become existential" in the Series F announcement.
The old sourcing move was to filter LinkedIn on "MLOps" or "LLMOps." That is now a generation behind. The right filters are GRPO, SFT, LoRA, RLHF, DPO in headlines, projects, or recent conference talks. Baseten's own acquisition of Parsed, a small RL startup focused on post-training and continual learning, confirms that "post-training engineer" is a distinct hiring pillar now, not a subspecialty of MLOps.
The valuation ramp cuts both ways
Baseten's own price chart is stark. Series D at $2.15B in September 2025. Series E at $5B in January 2026. Series F at up to $13B in June 2026. That is a greater than 6x valuation jump in about nine months. For a candidate walking in the door today, that ramp is a mixed pitch.
The upside: revenue grew roughly 20x year over year. Sacra estimates $600M annualized revenue in March 2026, up around 1,900% from $200M in December 2025. The platform now processes more than a billion inference calls per day across 87 clusters and 18 clouds. This is a real business with real load.
The downside for recruiting: new hires at $13B get materially worse strike prices than the engineers who joined at Series D. A Fireworks recruiter, sitting at a $552M valuation with comparable revenue trajectory, can pitch a much steeper equity upside curve to the same candidate. If you are recruiting for Baseten, you need to lean on scale and customer wins. If you are recruiting against Baseten, lean on the equity math.
The customer list leaked Baseten's shortlist. It also leaked its customers' shortlists to everyone else.