How Work Gets Done
AfterQuery reached a $3.2 billion valuation in September 2026, Forbes reported — a 10x increase in five months from its $300 million Series A valuation, Forbes's data shows — making it the fastest unicorn in Y Combinator history, per Forbes. The company reported $100M+ annual revenue run rate in April 2026, aitrainer.work's figures put it, with roughly 120 staff and a 100,000-expert network.
AfterQuery operates as two distinct organizations sharing a brand. One is a founder-led, in-person core moving at frontier-lab speed: high-agency, equity-aligned, with minimal management layers. The other is a marketplace of remote experts who do the actual data work: variable, project-based, paid hourly. Understanding which track you're joining, and how decisions flow in each, is the first filter for fit.
The founders — Spencer Mateega (CEO, 23) and Carlos Georgescu (CTO, 22), both Y Combinator Winter 2025 alumni — hold authority directly. BuiltIn's culture profile describes a "speed-and-results over structure" philosophy that "leans into ambition more than guardrails." Decisions happen in person and fast. The board's live data shows 11 salaried roles across engineering, research, and operations, with bands from $90,000 to $400,000 and a median around $200,000.
The in-office mandate codifies the rhythm. Five days a week in San Francisco, daily meals provided, commute stipends covered, Sunday night team dinners scheduled. BuiltIn notes this "codifies an on-site rhythm" where "employees make faster decisions and build tighter collaboration in person, though expectations skew toward presence and pace over remote flexibility." For the expert network — investment bankers, doctors, PhD researchers, software engineers recruited through a separate "Experts" platform — work is fully remote, asynchronous, and bursty. Projects launch in batches that fill in hours. Quiet stretches between projects are normal. Pay runs $35–$75 an hour for generalists, $75–$150 for technical specialists, $150–$250 for credentialed experts, paid weekly every Friday for the prior Monday–Sunday period. The platform guarantees it will never train on unpaid data; rejected tasks are excluded, not quietly shipped.
That dual track creates two different accountability models. Core employees get "end-to-end responsibility, practical projects, and autonomy to drive outcomes," per BuiltIn's culture summary, with "supportive leads and open-source opportunities" reinforcing ownership. Experts get per-task review cycles that can be slow or opaque — "slow or opaque review cycles and delayed approvals appear across projects, which can stall progress and hold up payments," the same source notes. Instructions shift mid-project as frontier labs refine what they're testing. The company directs all contractor inquiries to the Experts platform and discards general expert applications sent to the core hiring channel, a deliberate separation that clarifies pathways but also walls off the two populations.
The scoreboard is public and benchmark-driven. AfterQuery's own site leads with NVIDIA's public citation — the only data vendor named in Nemotron 3 Ultra's technical report, worth 11.4 GDPval points — and a 5x lift on Terminal-Bench 2.0. The $30 million Series A at $300 million valuation, BuiltIn found, and the $100 million revenue run rate appear in every hiring touchpoint. BuiltIn characterizes this as "Growth Milestone Broadcasting" that "sets a high-urgency, momentum narrative that attracts impact-seekers while calibrating employees to expect rapid change and ambitious targets." The pace is real. So is the process debt: organizational messiness, evolving processes, delayed or inconsistent responses creating uncertainty on expectations and timelines, weak management in review cycles particularly on the contractor side. The company is 18 months old. The operating practices are still being written.
The Upward Bet
AfterQuery's stated mission, "encode domain-specific excellence into forms that machines can learn," appears on the company's public site. The operating principle that separates AfterQuery from the data-labeling category is the upmarket bet. "AfterQuery's pitch was the opposite: go upmarket." That means recruiting specialists whose hourly rates make CFOs wince, then building tooling that makes each expert hour fifty times more productive. The customers are frontier foundation-model labs and a small set of large enterprises building agentic systems, buyers who can afford the output. The expert network has passed 100,000 vetted specialists across finance, law, medicine, software, and a long tail of niches that get harder to staff every quarter.
This principle shows up in product design. Four products, one job: get the expert's brain into the model. Supervised fine-tuning data consists of prompt-response pairs and step-by-step reasoning traces written by people who actually do the work: less "label the image," more "show your work." Reinforcement learning environments pair expert-designed prompts with grading frameworks so the model attempts, the rubric scores, and the policy improves, repeated ten million times. Custom agent environments are sandboxes wired to real APIs and tools so agents train on jobs that look like jobs, not toy chess. Computer use trajectories capture humans demonstrating the unglamorous reality of getting things done in software, click by click.
The founders frame this as a structural bet: the most valuable input to a frontier model is no longer compute or tokens, but the captured judgment of people who have done the work. "You can't scrape what a great lawyer notices in the eighth paragraph," the company's site states. "You have to ask her."
Customer engagement follows the same pattern. The company moves into workflows for insurance companies, publishers, healthcare companies, hedge funds, domains where generic tooling fails. The goal is to become a partner on the path toward AGI, not a vendor selling a dataset.
Revenue focus is explicit. The company disclosed $100M+ ARR in April 2026 and Forbes reported the company is already profitable. The founders position competitors politely, including Scale AI, Surge, Mercor, Invisible, Snorkel, and Labelbox, but argue the differentiation is the rubric: expert-designed grading frameworks rather than crowd-sourced labels. The bet is that if the next decade of AI runs on expert reasoning, somebody has to house the experts. AfterQuery is betting on its database of expert-graded examples.
What the Hiring Bar Selects For
The interview data paints a precise picture. Across six reported loops, every single candidate faced questions on AI evaluation systems. Ninety-seven percent hit data pipelines, both real-world and synthetic. Distributed systems appeared in 95 percent of loops. Reinforcement learning and RLHF showed up in 92 and 90 percent respectively. This is not a generalist screen; it is a filter for engineers who have already built the infrastructure that makes model training reliable at scale.
| Skill Area | Loop Coverage |
|---|---|
| AI Evaluation Systems | 100% |
| Data Pipelines (Real-world & Synthetic) | 97% |
| Distributed Systems | 95% |
| Reinforcement Learning (RL) | 92% |
| RLHF | 90% |
| High-throughput Asynchronous Processing | 87% |
| Python | 85% |
| Cloud Infrastructure (AWS/GCP) | 80% |
| Observability | 78% |
| Fault Tolerance & Resilience | 75% |
| Production Reliability / Reliability Engineering | 73% |
| RLVR | 70% |
Source: dataford.io interview guides, aggregated from reported loops
The process itself is lean. Interns get a recruiter screen followed by one to two technical rounds focused on data structures and algorithms. Senior roles run deeper into the systems above. Applications land in Ashby and are reviewed in arrival order, with no referral fast lane and no committee veto.
Compensation reinforces the signal. The board's live postings show those same aggregate bands with a $200k median across 11 salaried roles.
"A fast-paced, high-ownership environment that values direct feedback and high standards." (Scoutify company profile, May 2026)
The interview sentiment data is the clearest cultural tell. Zero percent positive. Thirty-three percent neutral. Sixty-seven percent negative. The two attributed comments capture the tension: "Afterquery offers excellent opportunities for engaging in open source projects" and "Project reviews can be time-consuming, which may impact workflow."
The company's client list tells you what the work actually is. The Nemotron ablation showed an 11.4-point lift on GDPval. Legora improved output quality 5 percent in one month after benchmark feedback. Thinking Machines Lab and select Chinese AI labs are also customers. The product is not labeling; it is verified reasoning trajectories from domain experts, calibrated by software, fed back into post-training pipelines. The hiring bar selects for people who understand that loop end to end.
Intern roles signal the pipeline. Open postings include Research Scientist (ICML), Research Scientist (Frontier Data), and Research Scientist (Post Training) alongside the standard Software Engineer Intern. The company hires researchers who can write production code and engineers who can design evaluation protocols.
With that YC speed record secured, the company founded by two 22-year-olds, profitable at $100M+ ARR, does not need to convince anyone to join. It needs to filter for the few who can operate without guardrails. The hiring bar is the culture.
What the Reviews Reveal
The public record on AfterQuery's workplace experience is thin but consistent across the few surfaces that exist. As of August 2026, Glassdoor showed eight reviews averaging 3.6 out of 5, with 100 percent of respondents saying they would recommend the company to a friend. Trustpilot, which tends to capture the platform's expert-contributor side more than full-time staff, sat at 3.3 out of 5 across 133 reviews. The Self Made Success review that surveyed both surfaces in August 2026 noted that Reddit feedback follows the same pattern: a split between people who have been paid on time and found the work engaging, and people stuck in long application queues, stalled task reviews, or projects that go quiet after onboarding.
Positive mentions cluster around three things: legitimate payments that actually arrive via Stripe on the weekly cadence the FAQ promises, intellectually interesting work that uses genuine professional expertise, and a platform that functions well enough to stay out of the way. Contributors who land projects matched to their existing credentials, such as medical doctors reviewing clinical summaries, engineers evaluating code outputs, and finance professionals stress-testing model reasoning, describe the tasks as manageable because the domain knowledge is already theirs. The pay transparency gets specific credit: most active listings show a full range rather than an "up to" ceiling.
The criticism side is equally specific. Slow application and task reviews appear in nearly every negative account; some contributors report pending applications for weeks, others describe approved work sitting in review limbo, delaying when earnings become payable. Project availability is inconsistent: the marketplace showed 11 pages of listings (well over 100 opportunities) in August 2026, but multiple reviewers note long gaps between assignments, paused projects, or applications that receive no response at all. Communication is described as weak or absent when things stall. Several accounts emphasize that "flexible" does not mean "low hours": the arcade/gaming expert listing asked for 40 hours per week, the medical MD role 20 hours, the general expert roughly 10 hours, and that the work demands professional-grade judgment, not beginner-friendly microtasks.
The same board data shows 11 salaried roles posted at those levels (median $200,000), a distinct population from the platform's contract experts. The Glassdoor sample of eight reviews is too small to separate full-time employee sentiment from contributor sentiment cleanly. The Self Made Success review's overall weighted score of 59 percent (17.5/30) reflects that same tension: credible platform, real payments, genuine Y Combinator backing, but high qualification bars, heavy time commitments on some roles, and inconsistent project flow that makes it unreliable as primary income.
Who Stays, Who Leaves
The split is sharp. AfterQuery runs two distinct labor pools simultaneously: salaried corporate roles in San Francisco and a much larger remote expert-contributor network. They operate on different contracts, different economics, and different cultural contracts.
The Corporate Track
The six salaried roles listed on Zero G Talent's board as of mid-2026 include Strategic Projects Lead ($200–400k), Technical Strategic Projects Lead ($200–400k), Senior Software Engineer, Infrastructure & Platform ($220–280k), Marketing Lead ($160–220k), Software Engineer - Applied AI / Platform ($140–200k), and Strategic Projects Associate ($200k); they share a common denominator. The titles signal ownership of outcomes, not execution of tickets. The three co-founders, Spencer Mateega (CEO, ex-Silver Lake, Morgan Stanley, Meta, Google), Carlos Georgescu (ex-Citadel Securities, Meta, Google), and Danny Tang, set the bar by reference. They have operated at the intersection of finance, AI, and high-stakes execution.
The compensation structure reinforces the signal. Base salaries run $140–400k with meaningful equity, a founder-led, pre-IPO cap table where the Series A ($30M at $300M valuation, April 2026, led by Altos Ventures with The Raine Group, Y Combinator, and BoxGroup) implies real upside if the run rate holds.
The Expert Network: Piecework at the Frontier
The remote contributor side, nearly 100,000 verified professionals across finance, law, medicine, engineering, software, ML/data science, business, and health, operates on a fundamentally different model. Pay ranges $40–50/hr for general expert work and up to $200–250/hr for top-of-skill specialists, per-project, paid via Stripe. The hiring funnel: apply at experts.afterquery.com/apply (1 day), vetting & domain assessment (5 days), onboarding (3 days), project matching (5 days), get paid (7 days). No equity. No benefits. No career path inside the company.
The Reddit thread from September 2024, the earliest public account of the contractor experience, is unsparing. "I don't feel like I'm hired as a SWE but like a DoorDash driver." The poster describes cloning random open-source repos, fixing issues, submitting Docker images, earning $15–150 per accepted task. "You're providing training data for AI." Another commenter: "sounds like another outlier.ai (or vosyn), I won't call it a scam, but really not worth it." A third: "I don't think this will help build your resume. You'd be better off with a side project." The thread's consensus: the work is real, the pay arrives, but the role is not a software engineering job. It is data labor: high-skill, domain-specific, but transactional.
Who thrives on this side: domain experts (physicians, attorneys, senior engineers, ML practitioners) who want flexible, well-compensated side income and understand they are generating training signal for frontier models. They treat it as a market, not a career. Who burns out: early-career developers expecting mentorship, resume lines that read "Software Engineer at AfterQuery," or a path to full-time conversion. The company has never promised conversion. The corporate roles recruit separately, with 18 open roles in San Francisco, mostly onsite, as of the 2026 careers page. The two tracks do not merge.
The Shared Burnout Profile
$100M+ ARR in 16 months with 11 salaried roles on the board implies extreme leverage per employee. That leverage comes from the expert network and the platform that routes work to it. The salaried team builds and operates the platform; the expert network supplies the cognition. Both sides require the ability to define the problem, execute the solution, and verify the result without a checkpoint meeting. The people who stay are the ones who already work that way. The people who leave are the ones who discover, too late, that no one is coming to tell them what to do next.
Working in AI? Zero G Talent tracks the openings: see every open AfterQuery role, browse AI jobs, the companies hiring, and the people building the field.