AfterQuery Reaches $3.2B Valuation as Frontier Labs Pay for Expert Reasoning Data

Eighteen months after launching in Y Combinator’s Winter 2025 cohort, AfterQuery reached a reported $3.2 billion valuation, according to Startup Fortune and Forbes. The valuation jump—up tenfold from a $30 million Series A round at a $300 million valuation in April 2026—makes it the fastest company in Y Combinator history to reach unicorn status.

The San Francisco-based company operates as an applied research lab. Rather than running traditional data labeling farms, AfterQuery structures complex human thought processes into training datasets, evaluation benchmarks, and reinforcement learning environments for frontier artificial intelligence labs.

The startup raised $30.5 million in total funding from investors including Altos Ventures, The Raine Group, BoxGroup, Latitude Capital, and Y Combinator, alongside angel investors from Google DeepMind, OpenAI, Anthropic, and Meta. CEO Spencer Mateega posted on X that annual recurring revenue reached several times the $100 million figure reported in April 2026, driven by enterprise contracts with model builders.

Engineering professional workflows into training environments

Foundation model builders have largely consumed public internet text and hit performance limits on synthetic data. To push models into professional knowledge work, labs require demonstration trajectories showing how practitioners evaluate tradeoffs, handle exceptions, and navigate desktop software.

AfterQuery focuses on five core technical pillars:

• Developing computer-use environments that train software agents to navigate application workflows and edge cases.
• Building multimodal datasets pairing text with interface layouts, financial tables, and technical diagrams.
• Constructing adversarial scenario data and guardrail suites with security specialists.
• Designing curation protocols to capture human judgment.
• Building evaluation benchmarks that test whether a model's internal reasoning steps match expert mental models.

The lab builds tool-calling reinforcement learning environments with real APIs and Model Context Protocol (MCP) servers, allowing software agents to practice chain-of-action tasks and error recovery. It also supplies supervised fine-tuning (SFT) datasets containing complex prompt-response pairs and reasoning traces, alongside rubric- and verifier-based reinforcement learning pipelines.

These products are already embedded in production pipelines. A technical report from Nvidia showed that workplace task data provided by AfterQuery was used during the training of its Nemotron 3 Ultra model, lifting its score on the real-world GDPval benchmark from 35.3 to 46.7 points. Beyond Nvidia, reporting by AI Magazine and 36Kr names Thinking Machines Lab, legal AI developer Legora, Alibaba, and Ant Group as partners and data clients. For Legora, AfterQuery built BAR, an agent evaluation benchmark spanning 5,161 cases and 11,075 documents across 28 legal practice areas.

Data generation runs through AfterQuery Experts, a network of roughly 100,000 verified practitioners in engineering, medicine, law, and finance. A profile by 34th Street Magazine detailed how the network also deploys university students to craft domain-specific problems, generate multi-step deliverables, and audit automated model answers against structured rubrics.

Founders with financial and engineering backgrounds

AfterQuery was founded in San Francisco by three operators with backgrounds in software engineering, trading, and institutional finance:

• Spencer Mateega, CEO, completed a B.S. in Finance and Statistics at Wharton and an M.S. in Computer Science at the University of Pennsylvania. He previously interned in technology private equity at Silver Lake, technology investment banking at Morgan Stanley, and software engineering at Meta and Google.
• Carlos Georgescu, CTO, studied Computer Science at the University of British Columbia, interned twice at Citadel Securities, and worked as a software engineer at Meta and Google. He previously founded and sold an education technology company before dropping out of university to launch AfterQuery.
• Danny Tang, Co-Founder, studied at Wharton and interned at Morgan Stanley, where he worked on the largest IPO of 2024. In high school, Tang and Mateega built and sold a startup together.

The founders initially entered Y Combinator's Winter 2025 batch aiming to develop automated financial agents. They shifted their focus after finding that commercial foundation models consistently failed on high-stakes white-collar tasks due to a lack of professional process data.

Hiring across systems infrastructure and project delivery

AfterQuery lists 20 employees on its core roster, and its open roles reflect an effort to scale both its data ingestion infrastructure and its client delivery teams.

On the engineering side, the company is hiring a Senior Software Engineer, Infrastructure & Platform ($220,000 - $280,000 USD per year) and a Software Engineer - Applied AI / Platform ($140,000 - $200,000 USD per year) to support sandbox execution, verifier pipelines, and post-training runs. To support the global pipeline of annotators and specialists, it is also recruiting a Software Engineer - Hiring Infra (LatAM) ($7,000 - $10,000 USD per month).

The largest cluster of open positions sits in project management and operations. AfterQuery relies on hybrid project teams to scope data contracts with foundation model labs and manage subject-matter workers. These roles span executive levels, such as the Technical Strategic Projects Lead ($200,000 - $400,000 USD per year), down through associate and intern levels.

Most roles are based onsite at the company's San Francisco headquarters, reflecting a dense in-person engineering culture.

Open roles at AfterQuery

See all open roles at AfterQuery