Sacra estimates that AfterQuery hit $100M in annualized revenue in April 2026.

Revenue is concentrated in a small number of very large accounts: every US-based frontier AI lab is a customer, alongside the world's largest technology companies, implying average contract values in the multi-million-dollar annual range rather than a broad SMB subscription base.

The revenue mix spans custom dataset engagements, bespoke RL environment development, off-the-shelf data packages, and enterprise consulting work.

Because deliverables are tied to specific model capability targets rather than seat licenses, contract values are driven by domain difficulty, environment complexity, and the depth of expert involvement required, which skews revenue toward larger, longer-duration engagements.

AfterQuery raised a $30M Series A at a $300M post-money valuation, announced on April 9, 2026. The round was led by Altos Ventures, with participation from The Raine Group, Y Combinator, Latitude Capital, BoxGroup, and angels from Google DeepMind, OpenAI, Anthropic, Meta Superintelligence Labs, and Microsoft AI.

Before the Series A, AfterQuery raised a $500K pre-seed round. The company is a Y Combinator W25 graduate, placing its founding and initial commercial launch in early 2025.

Total disclosed funding across all rounds is approximately $30.5M.

AfterQuery builds a data layer for training AI models to operate like domain experts rather than simply answer questions. Customers include frontier model labs and enterprises building advanced AI systems. The product combines human-curated training datasets, reinforcement-learning environments, and evaluation infrastructure intended to close the gap between a model that generates plausible outputs and one that behaves like a competent professional operator.

The core problem is that models trained on public internet data learn what answers look like, but not how expert work unfolds in practice: the sequencing of tool calls, judgment under ambiguity, and recovery from errors. In one example, before fine-tuning on AfterQuery data, a model handling a customer service workflow fabricated order IDs and placeholder emails. After training, it learned to first ask for identifying information, then use real values to call the right tools in the right order. That behavioral correction, rather than a benchmark score increase alone, is the product.

The offering has three layers. The first is supervised fine-tuning data: prompt-response pairs and reasoning traces that set baseline professional behavior before reinforcement learning begins. The second is RL environments: custom sandboxes built on top of real APIs, MCP servers, developer tools, browsers, and desktop interfaces where models learn to chain actions, call tools, and recover from failures inside realistic workflows. In one published example, a fresh Docker container boots, the model interacts via a terminal, a test suite runs, and the test results become the reward signal. The third is evaluation infrastructure: domain-specific benchmarks and rubric-based verifiers that test whether a model can do the work, not just pass a standard benchmark.

Public research artifacts, FinanceQA, IDE-Bench, Market-Bench, UI-Bench, App-Bench, VADER, and Terminal-Bench improvement work, show the playbook: identify a capability gap in a professional domain, build or align with a benchmark that exposes it, then create the training data and environments to close it.

On the supply side, the product relies on a network of nearly 100,000 verified practicing professionals across engineering, medicine, law, and finance. These contributors provide tacit judgment, domain fluency, and edge-case handling that synthetic data cannot replicate at the quality frontier labs require.

AfterQuery sells B2B to frontier model labs and large enterprises through direct enterprise sales. Buyers are typically post-training teams, eval teams, or enterprise AI groups with a specific capability gap to close, and engagements are consultative rather than self-serve: diagnose the failure mode, define the target workflow, deliver a custom or off-the-shelf dataset, environment, or eval package, then expand into more domains as results accumulate.

Monetization is project-based and programmatic rather than pure seat-license SaaS. Pricing is scoped to domain difficulty, expert mix, environment complexity, and dataset volume, which means the largest contracts are effectively outcome-oriented engagements where AfterQuery is paid to improve a model's performance on a specific professional task. Off-the-shelf datasets and reusable eval packages sit alongside custom work as a more scalable revenue layer.

The operating model is software-first and in-house rather than marketplace-style outsourcing. AfterQuery custom-builds the tools and workflows for each project and manages data creation internally, which raises fixed costs relative to pure labor-arbitrage vendors but gives it tighter quality control and more repeatable output. Over time, bespoke engagements generate reusable rubrics, environment templates, and domain playbooks that improve margins on subsequent work in the same vertical.

Its research-driven demand engine shapes go-to-market. By publishing benchmarks and case studies that show where models fail and how curated data fixes those failures, AfterQuery uses public research as sales enablement, proof of competence that reduces friction with technically demanding buyers at frontier labs. ISO 27001 and SOC 2 Type II compliance act as deal enablers for enterprise accounts that need to share proprietary workflows, codebases, or sensitive domain data, where compliance is a procurement gate rather than basic hygiene.

The post-training data market has shifted from generic annotation toward expert-generated reasoning data, RL environments, and domain-specific evaluation infrastructure. That shift has pulled a larger set of well-capitalized competitors into AfterQuery's core market.

Scale AI is the clearest competitive benchmark. Its GenAI Data Engine combines vetted subject-matter experts, RLHF pipelines, and model evaluation tooling, and Scale launched dedicated RL Environments in early 2026, with nearly half of new data-training projects now involving reinforcement-learning environments.

Scale's advantage is breadth and procurement familiarity: it can bundle data creation, infrastructure, and compliance into a single vendor relationship that enterprise buyers may find easier to manage. AfterQuery's counter is depth, with a more research-native process that starts with failure-mode analysis rather than fulfillment.

Mercor and Surge AI are the most direct strategic peers. Mercor's research arm sells benchmarks, eval environments, RL environments, and large-scale human datasets, claiming penetration into the top five AI labs and six of the Mag 7, a customer profile close to AfterQuery's. Surge AI competes in expert professional domains and RLHF, and uses a public research and leaderboard presence to influence benchmark selection upstream of a data sale.

Labelbox has moved into the category with RL data, private AGI benchmarks, arena evals, and an expert network through its Alignerr platform, claiming partnerships with over 80% of leading US AI labs and access to more than 1.5 million knowledge workers, including tens of thousands of PhDs and licensed professionals. Turing approaches the market from a talent-marketplace base, offering frontier data packs for coding, STEM, multimodality, and domain-specific reasoning alongside RL gyms and human-in-the-loop synthetic data pipelines. That makes it a credible alternative when customers prioritize throughput and coverage over boutique research depth.

The most important indirect threat is frontier labs building internal post-training capability. OpenAI and Meta are both actively hiring for post-training pipelines, which can shift a vendor from core partner to overflow and niche-domain supplier as labs mature.

A second pressure comes from hybrid synthetic pipelines, where competitors use experts to design rubrics and audit outputs rather than generate every example from scratch, compressing cost in ways that could erode AfterQuery's margin advantage on high-volume work. Eval-tooling platforms like Braintrust, LangSmith, and Patronus are an adjacent threat as well, because they let customers run regression datasets and organize human review in-house, reducing the need to outsource custom validation work.

AfterQuery's current revenue base is anchored in frontier model labs, but the post-training stack it has built applies to a broader set of buyers and use cases as agent deployment diffuses through the economy. The main expansion path is into enterprises moving from AI experimentation to production deployment.

The clearest near-term expansion is from model labs into enterprises building and deploying agents. McKinsey data shows that roughly 62% of organizations are experimenting with AI agents while only 23% report scaling one anywhere in the enterprise, a gap that creates demand for workflow grounding, domain data, and deployment infrastructure of the type AfterQuery provides.

The Raine Group partnership shows the motion: AfterQuery enters through a knowledge-first diagnostic of internal workflows, identifies where firm-specific expertise is bottlenecked, and builds the data and agent infrastructure around it. That approach is repeatable across finance, legal, healthcare, and professional services firms with the budget and workflow complexity to justify a high-touch engagement.

AfterQuery already has footholds in finance, software engineering, medicine, and law, and each vertical has room for a fuller stack of benchmarks, datasets, rubrics, simulators, and deployment services. FinanceQA and the Raine Group work give the company a base to expand from analyst reasoning into adjacent workflows such as diligence, compliance review, and valuation QA.

Software engineering is another expansion lane. IDE-Bench, Terminal-Bench improvement work, browser and desktop environments, and code-generation products give AfterQuery exposure to coding agents, devtools vendors, and enterprise engineering copilots, not just text reasoning models. Healthcare and other multimodal domains expand TAM because customers need safer reasoning and workflow simulation under stricter standards, not only better answers.

The spread of MCP as an open standard for connecting AI applications to external systems broadens the surface area for AfterQuery's tool-calling RL environments. As more agents are built on top of APIs, MCP servers, and enterprise software, the need for training and evaluation in tool-rich environments grows, and AfterQuery's existing product surface maps to that demand.

This creates a path toward becoming a cross-platform agent-readiness layer: helping customers tune and validate models across APIs, internal databases, developer tools, and business software rather than serving only frontier labs on capability benchmarks. Partnerships with systems integrators and cloud providers could accelerate distribution into large enterprises that have agent budgets but lack internal post-training expertise.

Lab concentration: AfterQuery's revenue is concentrated in a small number of frontier AI labs that are both its largest customers and potential competitors, so a decision by even one or two labs to internalize post-training data operations or shift to a bundled platform vendor could create a material revenue gap that the enterprise channel would take time to fill.

Margin structure: AfterQuery's quality differentiation depends on verified domain experts, custom-built tooling, and in-house workflow management rather than marketplace-style labor arbitrage, so its cost structure is higher than competitors like Turing or Mercor that can flex supply more cheaply, which constrains gross margin expansion unless the company converts bespoke engagement work into reusable off-the-shelf inventory at scale.

Benchmark commoditization: AfterQuery's research-driven demand engine depends on its benchmarks and evals being treated as authoritative measures of real professional capability, but as Scale AI, Labelbox, Surge AI, and Mercor invest in their own public evaluation infrastructure, the benchmark landscape could fragment and dilute AfterQuery's ability to use published research as a durable sales and credibility moat.