AfterQuery
AfterQuery is an applied AI research lab that builds expert-level training datasets, SFT traces, and RL environments for frontier foundation model developers and large enterprises with proprietary institutional knowledge. It serves AI labs and enterprises via off-the-shelf datasets, custom data production, and end-to-end AI implementation.
- Company typePrivate
- Founded2025
- HeadquartersSan Francisco, United States
- Headcount51–100
- GTM typeB2B
- OfferingSoftware
What AfterQuery does
AfterQuery is an applied AI research lab that produces expert-level training data for foundation model developers. The company curates datasets across four product lines: Supervised Fine-Tuning traces, Reinforcement Learning environments with expert-designed rubrics, Agent Environments (API and MCP-compatible), and Computer Use Trajectories. It serves two customer segments: frontier AI research labs (the company claims to power every major frontier lab) and large enterprises with proprietary institutional knowledge (e.g., The Raine Group in investment banking). Its underlying technology stack includes AfterQuery Atlas, an in-house annotation platform that ingests enterprise document corpora into searchable vector embeddings, and a data labeling agent harness that converts institutional precedents into agent-ready data.
The business operates a hybrid GTM: enterprise field sales with forward-deployed engineers who co-build onsite at client offices, paired with a self-serve request flow for off-the-shelf datasets via the company website. Off-the-shelf offerings include Terminal-Bench (terminal agents), τ²-bench (customer service agents), GDPval (professional knowledge work), Office Agent, and SWE agent datasets, all sourced from a network of nearly 100,000 domain professionals. Revenue is generated through quote-based custom dataset commissions, off-the-shelf dataset licensing, and end-to-end enterprise AI implementation engagements. NVIDIA has publicly cited AfterQuery's GDPval training dataset in its Nemotron 3 Ultra technical report, and AfterQuery has disclosed surpassing $100 million in annual revenue run rate alongside its April 2026 $30M Series A at a $300M valuation led by Altos Ventures.
Founding backgrounds span Goldman Sachs, McKinsey, Jane Street, Palantir, NVIDIA, and Google. The company is headquartered in San Francisco with 51-100 employees and 25 open roles as of source date. AfterQuery Inc. is a Delaware-incorporated private company founded in January 2025, with CEO Spencer Mateega and co-founders Carlos Georgescu (CTO) and Danny Tang.
AfterQuery firmographics
Firmographics- Name
- AfterQuery
- Legal name
- AfterQuery Inc.
- Website
- https://afterquery.com
- Company type
- Private
- Founded year
- 2025
- Operating status
- Operating
- Headcount range
- 51–100 employees
- Short description
- AfterQuery is an applied AI research lab that builds expert-level training datasets, SFT traces, and RL environments for frontier foundation model developers and large enterprises with proprietary institutional knowledge. It serves AI labs and enterprises via off-the-shelf datasets, custom data production, and end-to-end AI implementation.
- Ownership category
- akta.pro rank
AfterQuery industry classification
Industry- Product category
- AI Training Data
- NAICS
- Computer Systems Design and Related Services (54151)
- SIC
- Services-Prepackaged Software (7372), Services-Computer Programming Services (7371)
- akta.pro primary industry
- Data Quality, Profiling & Validation (within Integration) (HDAEACAI)
- akta.pro secondary industry
- Skills, Competency & Capability Management (Skills Frameworks, Badging within LMS) (EDAFAAAE)
Keywords
Where AfterQuery is headquartered
LocationHeadquarters
- HQ city
- San Francisco
- HQ country
- United States
- HQ region
- North America
Offices1 record
Markets served
AfterQuery business model
Business model- GTM type
- B2B
- Offering type
- Software
- Cost components
- Personnel, Technology or R&D, Operations, Marketing or Sales, Infrastructure
Revenue model
- Off-the-shelf dataset sales: Pre-built, commercially available training datasets (τ², Terminal-Bench, GDPval, Office Agent, SWE agent) sold directly to AI labs and enterprise customers via the website's 'Browse our Off-the-shelf Datasets' option.
- Custom dataset production: Bespoke datasets commissioned by customers via the 'Request a Custom Dataset' contact form; engagements scoped per project with expert labor priced per dataset, supporting frontier lab post-training needs.
- Enterprise AI consulting & implementation: Forward-deployed consulting and end-to-end enterprise AI implementation engagements, including firm-specific data system builds (e.g., Raine Search for The Raine Group) that integrate data infrastructure with frontier model deployments.
- Expert network labor payments to contributors: Inputs sourced from a network of nearly 100,000 professionals across multiple sectors who are paid for expert labeling, trajectory generation, and rubric design feeding the data products.
Pricing tiers
| Model | Billing | Price |
|---|---|---|
| Other | Multi-year contract | Custom datasets and enterprise AI consulting: quote-based |
| Other | Pay-as-you-go | Off-the-shelf datasets: quote-based |
Go-to-market motion3 records
Distribution channels3 records
Marketing channels6 records
AfterQuery product offering
Product offeringCore offering
AfterQuery is an applied research lab that curates expert-level AI training data for frontier foundation model development. The company sells Supervised Fine-Tuning (SFT) traces, Reinforcement Learning environments with expert-designed rubrics, Agent Environments (API/MCP), and Computer Use Trajectories, delivered as off-the-shelf datasets (Terminal-Bench, τ²-bench, GDPval, Office Agent, SWE agent) to frontier AI labs and as custom datasets and enterprise AI implementations to large organizations with proprietary institutional knowledge.
Product overview
AfterQuery is an applied research lab offering a platform of data products for AI model development rather than a single unified product. Its core offerings fall into four categories as described on the homepage: Supervised Fine-Tuning (SFT) data, Reinforcement Learning + Rubrics, Agent Environments (API / MCP), and Computer Use Trajectories. These are delivered through off-the-shelf datasets (Terminal-Bench, SWE agent, τ²-bench, GDPval, Office Agent Training, and agentic post-training datasets), custom datasets, and enterprise AI consulting/implementation services. Underpinning the platform is AfterQuery Atlas, an internal annotation platform that powers tools like Raine Search. The off-the-shelf datasets are the primary commercial products, while the enterprise AI solutions team delivers custom data and end-to-end implementations such as the Raine Search deployment for The Raine Group.
Differentiator
Problem solved
Functional benefit
Products and services
- Supervised Fine-Tuning (SFT) Data High-quality prompt-response pairs and chain-of-thought reasoning traces that teach models how to behave across complex tasks; provided as off-the-shelf and custom datasets for AI labs and enterprise customers.
- Reinforcement Learning + Rubrics Expert-designed prompts paired with grading frameworks for reasoning and code generation, turning subjective expert judgment into scalable reward signals for RL training.
- Agent Environments (API / MCP) Custom environments across APIs, tools, and services that enable training and evaluation of agents in real workflows, with API and Model Context Protocol (MCP) integration support.
- Computer Use Trajectories Human-demonstrated interactions across browser and desktop environments that teach models to navigate and operate software end-to-end.
- Terminal-Bench Training Dataset Curated dataset of expert-labeled terminal-agent trajectories for software engineering, system administration, and data processing tasks; used to improve openai/gpt-oss-20b from 3.1% to 17.0% on Terminal-Bench 2.0.
- τ²-bench Training Dataset Dataset of customer-service conversational tasks across airline, retail, telecom, banking, and other domains; used to fine-tune Llama-3.1-8B-Instruct and improve performance up to 4.33x in retail domains.
- Office Agent Training Dataset Dataset of 800 professional-work tasks with well-seeded workspaces, messy inputs, specific deliverables, and realistic constraints; cited by NVIDIA in the Nemotron 3 Ultra technical report.
- GDPval Training Dataset Dataset of professional knowledge-work tasks used for on-policy distillation training, achieving a +21.4% net win-loss margin on the GDPval benchmark.
- SWE Agent and Agentic Post-Training Datasets Additional off-the-shelf datasets for software engineering agent training and general agentic post-training, accessed via the contact form.
- Custom Datasets Bespoke training data solutions built per customer requirements, available through the sales inquiry channel with engagement-specific pricing negotiated via separate agreements or order forms.
- Enterprise AI Consulting Advisory engagement that helps enterprises understand where AI systems break when meeting real workflows, messy data, and organizational context.
- End-to-End Enterprise AI Implementation Full implementation service that builds AI systems tailored to a customer's workflows, including forward-deployed engineering and integration into existing enterprise systems, as exemplified by the Raine Search deployment.
Quantifiable outcome
- Improved openai/gpt-oss-20b on Terminal-Bench 2.0 from 3.1% to 17.0% (SFT + RLVR pipeline), beating Gemini 2.5 Flash without training on a single official eval task
- +3 more outcomes
Companies that use AfterQuery
Customer profileNamed customers3 records
Segments3 records
Ideal customer profiles3 records
AfterQuery technology and API
TechnologyTechnology focussed Yes
API detail
- Has API
- Yes
- API docs
- API detail
Core technology
AI maturity
App detail
AI capability10 records
Feature6 records
AfterQuery partnerships and signals
Strategic signalPartnerships
Five partnerships are on record, tiered core.
- The Raine GroupcoreStrategic partner and co-development customer: AfterQuery co-built Raine Search, a firm-specific semantic search tool running on AfterQuery Atlas, by deploying forward-deployed engineers into Raine's Midtown New York offices; The Raine Group also participated as an investor in the $30M Series A.
- Thinking Machines (Tinker)coreAfterQuery's training pipeline uses Thinking Machines' Tinker SDK for both supervised fine-tuning and reinforcement-learning stages, including on-policy distillation experiments referenced in the research blog.
- Harbor FrameworkcoreAfterQuery uses the open-source Harbor framework to build multi-turn RL environments for terminal-agent tasks and for Terminal-Bench 2.0 evaluation.
- NVIDIAcoreAfterQuery's Off-the-shelf GDPval Training Dataset was used by NVIDIA in development of Nemotron 3 Ultra (cited in NVIDIA's technical report); AfterQuery researchers also ran on-policy distillation experiments on NVIDIA Nemotron-3-Nano and Nemotron-3-Super models.
- Experts.afterquery.com Contributor NetworkcoreAfterQuery maintains a separate Experts marketplace (experts.afterquery.com) for engaging expert contributors on contract jobs to produce training data; nearly 100,000 professionals participate across multiple sectors.
Scale indicators10 records
Recent moves6 records
Expansion highlights6 records
AfterQuery competitors and assessment
Company assessmentDirect peers
- Scale AI: Scale AI is the largest commercial provider of training data for AI models, offering data labeling, RLHF, evaluation, and frontier-model testing services. It is the most direct competitor to AfterQuery across both off-the-shelf datasets and enterprise AI deployments.
- Surge AI: Surge AI provides high-quality human-labeled data for frontier AI labs with a similar quality-and-expert-focus positioning to AfterQuery. Both companies sell expert-produced SFT/RL data to the same frontier lab buyer set.
- Snorkel AI: Snorkel AI offers a data-centric AI platform combining programmatic labeling, expert annotation, and synthetic data for enterprise model development. It overlaps with AfterQuery's Atlas annotation platform and custom-dataset production business.
- Labelbox: Labelbox provides an annotation platform plus managed labeling services for AI training data. Its platform-plus-services model closely mirrors AfterQuery's Atlas platform and custom-dataset offering for enterprise buyers.
Broad incumbents
- Appen: Appen is a long-standing incumbent in data annotation and collection for AI, serving large enterprise and government buyers globally. It is broader and less frontier-lab focused than AfterQuery but addresses the same core market of training-data production.
- TELUS Digital (formerly TELUS International AI): TELUS Digital is a large-scale data annotation and digital CX provider that serves enterprise AI buyers. It is broader and lower-margin than AfterQuery but operates in the same training-data production category.
- iMerit: iMerit provides expert-in-the-loop data labeling and annotation services across computer vision, NLP, and content safety. It competes with AfterQuery for skilled annotation labor and for enterprise AI buyers.
Emerging players
- Mercor: Mercor recruits domain experts to produce AI training data and has rapidly scaled via frontier lab contracts. It is a fast-emerging competitor in the same expert-data category AfterQuery leads in.
- Micro1: Micro1 builds expert-sourced training datasets and evaluation environments for frontier AI labs, with a focus on human expert labor at scale. It directly competes with AfterQuery for both expert contributors and frontier lab customers.
Others
- Handshake: Handshake operates a large professional talent network and has expanded into AI training data via partnerships with labs. It is an adjacent ecosystem participant — a contributor-network partner or potential competitor for expert labor — rather than a direct dataset seller.
Market position
Strengths5 records
Weaknesses5 records
Competitive moat5 records
Key risks7 records
Key highlights7 records
Customer concentration
AfterQuery social profiles
Digital presenceAfterQuery financial estimates
Financial estimateRevenue estimate
Valuation estimate
AfterQuery leadership team
Management profileNumber of profiles
Profiles3 records
AfterQuery funding detail
Funding detailFunding overview
Funding rounds3 records
Investors6 records
Funding detail is available on the Subscription and Enterprise plan.Contact sales →
AfterQuery M&A and investment
M&A and investmentM&A
Investments
M&A and investment is available on the Subscription and Enterprise plan.Contact sales →
Frequently asked questions about AfterQuery
What does AfterQuery do?
AfterQuery is an applied research lab that curates expert-level AI training data for frontier foundation model development. The company sells Supervised Fine-Tuning (SFT) traces, Reinforcement Learning environments with expert-designed rubrics, Agent Environments (API/MCP), and Computer Use Trajectories, delivered as off-the-shelf datasets (Terminal-Bench, τ²-bench, GDPval, Office Agent, SWE agent) to frontier AI labs and as custom datasets and enterprise AI implementations to large organizations with proprietary institutional knowledge.
Is AfterQuery a public or private company?
AfterQuery is a private company. It is classified as venture growth investor backed and is currently operating.
When was AfterQuery founded?
AfterQuery was founded in 2025. It employs 51 to 100 people.
Where is AfterQuery based?
AfterQuery is headquartered in San Francisco, United States, in the North America region.
How does AfterQuery make money?
Four revenue lines are on record. Off-the-shelf dataset sales are the primary driver. The others are custom dataset production, enterprise AI consulting & implementation and expert network labor payments to contributors.
Who are AfterQuery's main competitors?
Direct peers on record are Scale AI, Surge AI, Snorkel AI and Labelbox. Broad incumbents are Appen, TELUS Digital (formerly TELUS International AI) and iMerit. Emerging players are Mercor and Micro1. Handshake is listed as an others.
Does AfterQuery have an API?
Yes. AfterQuery offers API-accessible agent environments and MCP-compatible environments for training and evaluating AI agents. The platform exposes its annotation and data labeling capabilities via AfterQuery Atlas and its data labeling agent harness. The Terminal-Bench work uses a CLI-driven Tinker SDK where a single CLI call handles GPU allocation and training (e.g., python3 train_sft.py with model_name, learning_rate, batch_size, lora_rank flags). The Tau2 evaluation harness exposes agent/user LLM configuration via command-line arguments (--agent-llm, --user-llm, --api_base, --temperature). APIs are primarily partner/enterprise-facing for frontier AI labs and enterprise customers, not public self-serve.
What industry is AfterQuery in?
AfterQuery's product category is AI Training Data. Its primary akta.pro industry code is HDAEACAI, Data Quality, Profiling & Validation (within Integration), with a secondary code of EDAFAAAE, Skills, Competency & Capability Management (Skills Frameworks, Badging within LMS). Its NAICS code is 54151 and its SIC code is 7372.