Who Gets Hired and Onto Which Teams
Poolside AI has grown its engineering headcount 159 percent since 2023 to 264 employees, Revelio Labs reported, with every current opening designated remote across EMEA and the U.S. East Coast. The company hires like an enterprise infrastructure firm that happens to train its own foundation models — because that's what it is becoming.
Jason Warner and Eiso Kant founded the company in 2023. Warner oversaw GitHub Copilot's launch as CTO; Kant built code-search startup source{d} and engineering analytics firm Athenian. They stocked the early team with researchers and engineers from DeepMind, Yandex, Amazon, and Uber.
By March 2026, headcount reached roughly 264 — up from 102 in 2023 — with engineering growing at 18 percent year over year, the fastest of any function, Revelio Labs data shows. The workforce splits three ways: Finance and Operations holds the largest share at 47 percent, reflecting the operational weight required to run 10,000-GPU training clusters and enterprise deployments in regulated environments. Engineering sits at 33 percent and climbing. Sales and Marketing rounds out the remaining 20 percent, bolstered by the April 2024 hire of former GitHub Chief Revenue Officer Paul St. John to lead enterprise go-to-market. Geography mirrors the remote-first model: 40 percent of employees sit in North America, 19 percent in Western Europe, and 18 percent in Northern Europe.
Active job postings on the Zero G Talent board show the hiring mix in real time. All six current openings carry the "Member of Engineering" title across distinct specializations: Interfaces (Full Stack), Data & Analytics, Multimodality (Research Lead), Experiment Platform, Inference Infrastructure, and Infrastructure. Each listing specifies "Remote (EMEA/East Coast)" or "Remote (EMEA)", the company's shorthand for a hiring corridor spanning European time zones and the U.S. Eastern seaboard. That corridor aligns with where the existing team clusters and where enterprise customers operate.
Beyond the board listings, public role descriptions reveal two additional archetypes. A Member of Engineering (Post-training) builds data pipelines and environments for agentic use cases, researches and implements post-training algorithms, and designs experiments (work at the intersection of research and production infrastructure). Forward Deployed Research Engineers embed directly with customer teams to customize and align models to specific business logic and evaluation metrics. These engineers operate inside the client's environment (on-premise, VPC, or air-gapped), making the role part solutions architecture, part research, and part deployment engineering.
The product isn't publicly available as of May 2025; it's sold to Global 2000 companies and government agencies that demand on-premise installation, auditability, and compliance-grade governance. Engineers at Poolside ship artifacts that run in someone else's security boundary, often without direct access to the environment.
What It Pays
Poolside pays at the top of the private-market AI band.
| Role / Level | Base Salary Range (USD) | Total Comp Median (USD) | Equity Trend (YoY) |
|---|---|---|---|
| Member of Engineering (IC) | $150K–$450K* | $220K | +13% (mid) to +15% (senior) |
| Director of Efficiency | $230K (reported) | — | — |
| Senior Director (AI) | — | — | +40% |
* U.S. range per Mistral AI spokesperson to Business Insider (July 2025); excludes equity, signing bonus, and benefits. Poolside-specific base range not publicly disclosed.
The median total compensation for a Member of Engineering — Poolside's title for what most firms call a software engineer — sits at $220,000 per year, per Levels.fyi data collected September 2026. That figure blends base, bonus, and equity. The highest-reported U.K. package reaches £321,000 (roughly $410,000), while the U.K. median lands at £152,000 ($193,000). The spread reflects both the remote-first footprint in those regions and the premium attached to the "AI engineer" label in 2025–26.
For context, peer firms show wider bands: OpenAI lists Member of Technical Staff roles at $200,000–$530,000, Anthropic research engineers at $340,000–$690,000, and Mistral AI scientists at $280,000–$350,000. Poolside's median total comp of $220,000 trails the top of those ranges but exceeds the median at many Series B–C AI startups.
Equity is where the market has moved hardest. SaaStr's September 2025 analysis of private-company grant data — drawn from Carta and option-grant filings — shows Senior Director–level AI engineers seeing equity increases of 40 percent in under two years. The same dataset puts mid-level grants up 13 percent, senior ICs up 15 percent, managers up 20 percent, senior managers up 27 percent, and directors up 30 percent. Poolside, founded in 2023 and now at roughly 350 employees per Levels.fyi, sits squarely in the cohort granting those inflated packages. The company's job titles, including Member of Engineering (Interfaces – Full Stack), Member of Engineering (Inference Infrastructure), and Member of Engineering (Multimodality – Research Lead), map to the IC and research-lead levels where equity bumps of 13–15 percent are now baseline.
Dilution follows. SaaStr warns that AI startups "have less headcount on average than their pre-AI peers, but far more dilution." Every 40 percent equity bump at the Senior Director level expands the option pool and shrinks everyone else's slice. Poolside's estimated revenue of $10 million to $50 million (Levels.fyi, 2026) means the equity's paper value hinges on a future funding round or exit — not current cash flow.
The practical takeaway: a Member of Engineering offer today will likely carry a base in the low-$200,000s and an equity grant sized to the 13–15 percent YoY inflation benchmark. Negotiate the equity clawback structure (performance milestones tied to model-release or revenue targets) before you sign. The market price isn't normalizing; SaaStr's September 2025 verdict stands: "Every month you wait to hire critical AI talent, the market price goes up."
How the Hiring Process Works
The sequence: a screening call with an engineering lead, two technical interviews, then culture-fit conversations. That structure mirrors the remote-first, mixed engineering model the company advertises across its six open roles, all listed as remote in those time zones.
A Glassdoor reviewer described the process as "very research oriented" with "no leetcode-like coding questions." Technical interviews replace whiteboard drills with research discussions. Culture-fit calls assess self-direction, a trait the remote model demands.
AI-driven screening tools now rank applicants before a human sees them at many frontier labs. Platforms like TalentGPT score every sourced profile against ranking criteria and surface reasoning for each score, replacing the first two manual resume reviews. Candidates whose materials don't map cleanly to the posted requirements (missing keywords, unclear project scope, no quantified impact) may never reach the engineering lead screen.
Strong applications share three traits: a project narrative that mirrors the target track's technical language, evidence of end-to-end ownership (not just model training but deployment, monitoring, iteration), and concise communication. Candidates who treat the process as a research collaboration, not an exam, advance.
Where the Work Happens
Poolside was built remote-first by design. In a 2026 interview, Kant stated the company made "a very conscious decision" early on: "We're not gonna hire any researchers in the Bay Area. We're gonna look for talent everywhere else in the world." That stance remains visible in current openings. Zero G Talent's board lists every active engineering role under those same remote designations. No Bay Area location appears.
The team is small: fewer than 70 researchers and roughly 35 engineers as of mid-2026, per the same founder interview. That headcount runs a Model Factory shipping open-weight models such as Laguna S 2.1 (118 billion total parameters, 8 billion active) on eight-week cycles, with 10,000–20,000 experiments cut per month. Distributed coordination at that tempo requires deliberate infrastructure, not just a Slack workspace. The company's public posts describe a production-grade platform for running AI agents inside customer boundaries, with administrators defining access and every action visible, a system that itself reflects the internal tooling the distributed team relies on.
Physical footprint exists, but it is not a traditional headquarters. Project Horizon, an ambitious data-center campus in West Texas originally planned at 2 gigawatts, signaled an intent to secure industrial-scale compute independently. The company had a six-week window at the end of 2025 to raise $2 billion for a 40,000-GPU GB300 cluster slated to come online in January 2026. That plan shifted when Nvidia agreed to invest $1 billion and pay another $6 billion to license Poolside's technology, bringing more than 100 Poolside employees into Nvidia's own AI development efforts. The partnership gives Poolside "substantially greater access to capital, computing infrastructure and Nvidia's own hardware ecosystem," per USA Herald's August 2026 reporting. In practice, the West Texas campus is no longer the primary compute path; Nvidia's infrastructure and the Nemotron Coalition (shared expertise, data, and compute across Mistral, Thinking Machines Lab, Perplexity, and now Poolside) replace it.
For candidates, the "where" is two-layered. Day-to-day work happens wherever the engineer lives, aligned to EMEA or East Coast hours for collaboration overlap. The compute that matters (training runs, evaluation at scale, the Model Factory pipeline) runs on Nvidia-backed clusters accessed remotely. There is no relocation package to a campus because the campus that mattered got folded into a partner's fleet. The trade-off is clear: you join a sub-110-person research and engineering team with frontier compute access and open-weight shipping cadence, but you operate without a physical office anchor. The company's own phrasing, "You choose the model. You choose where it runs. You choose the infrastructure and deployment pattern that makes sense for each workload," applies to its own people as much as to customers.
Who Thrives Here
The signal from Poolside's own records (job postings, founder interviews, user feedback on its models) converges on a consistent profile. The people who last and advance share a cluster of traits that map directly to what the lab is building: production-grade, enterprise-deployable foundation models and agentic systems for software engineering.
Engineering bias as cultural default
Kant described the company's culture in a July 2026 Latent.Space interview as having "a very strong engineering bias" that "helped us get to where we were." That bias shows up in every hiring signal. The Zero G Talent board lists six current openings, all titled "Member of Engineering" across the same six specializations. No product managers, no designer roles, no sales engineers. The organization hires engineers to do engineering work, including the work of deciding what to build.
Outcome ownership over model shipping
Poolside's own product literature states "Outcome Ownership: Joint responsibility for adoption, impact, and business results rather than just shipping models." That line isn't marketing copy; it describes the behavioral filter. The Events Lead role (one of the few non-engineering positions publicly advertised) lists "Own the entire show," "Connect events to real outcomes," and "Own the money story" as responsibilities. The same language appears in how the lab evaluates model releases: developers testing the Laguna series note that the model "really shone in an adversarial review role" and produced code "minimal, in the style of the code base, and followed the coding standards very well." The lab rewards engineers who think like the enterprise customers who will run these models on-prem, in VPCs, with strict RBAC and audit requirements.
Security-first, enterprise-grade mindset
Every technical role at Poolside touches deployment surfaces that must satisfy enterprise security review. The Quasa review highlights "Enterprise-First Design: On-prem, VPC, or workstation deployment with strict data privacy, role-based access control, governance, and auditability" as a core pillar. Engineers who thrive here internalize those constraints early. They don't treat security as a compliance checkbox; they design for it because the product's value proposition depends on it. The same developer feedback that praised Laguna's code quality also flagged "occasional looping issues" and reasoning that stopped "at around 70,000 tokens in long context tasks", bugs that would be unacceptable in a customer's air-gapped environment. The engineers who stay are the ones who treat those reports as their problem, not QA's.
Self-directed execution in a remote-first lab
All six current engineering postings are remote (EMEA/East Coast). There is no office mandate to create structure. The people who succeed build their own: they define the experiment, instrument the platform, and push the inference stack without a stand-up telling them to. The Fern Labs acquisition announcement (November 2025) framed the deal as "transforming enterprises into AI-native, agentic organizations", language that implies autonomy at every level. Candidates who need daily syncs or detailed tickets to move forward don't match the velocity the lab expects.
Thoroughness over speed
Developers consistently describe Poolside's models as "very thorough when researching, and reasoning through solutions" and note "I'll use this one when I want the job done well, not quickly." That preference mirrors the hiring bar. The lab's $500 million pre-product raise (Deloitte reported in November 2024) bought runway to solve hard problems correctly, not to ship demos. People who optimize for demo velocity leave; people who optimize for correctness in long-horizon agentic loops stay.
Mission alignment that survives the hype cycle
"Poolside exists to be this company: to build a world where AI will be the engine behind economically valuable work and scientific progress." That mission statement, from the company's own site, filters for a specific kind of ambition. The SaaStr commentary on Poolside's fundraising ("A $9B outcome sometimes doesn't clear the bar for seed investing in 2026") reveals the external pressure: the market expects category-defining outcomes. Employees who treat this as a resume builder for the next AI wave churn out. The ones who remain are building the infrastructure they believe will still matter in 2030.
The hiring profile
Put it together: the lab selects for senior engineers who have shipped production ML systems, prefer designing the experiment platform to running the experiment, write code that passes an adversarial review, treat security requirements as design inputs, work autonomously across time zones, and measure progress in adopted capability, not merged PRs. That's who clears the screen. The rest of the industry is hiring for the same title; Poolside is hiring for a narrower job.
Working in frontier tech? Zero G Talent tracks the openings: see every open Poolside AI role, browse frontier tech jobs, the companies hiring, and the people building the field.