Applied Compute
NativePrivateUpdated November 3, 2026AI model post-training/RL infra, ex-OpenAI founders
- Website
- www.appliedcompute.com
- CEO
- Yash Patil
- Founded
- 2025
- HQ
- San Francisco, United States
- Total Funding
- $160M
- Employees
- 11-50
Products & Offerings
Specific Intelligence custom agents and models
A forward-deployed service that trains custom models with reinforcement learning on a company's own data and runs them as agents for specific workflows, such as the merchant onboarding agent built for DoorDash.
AC2 (Applied Compute Agent Cloud)
A training and inference cloud for open-weight models that lets teams bring their own harness, post-train with RL, serve the models on dedicated inference and keep learning from production traces.
Executive Summary
What they do
- Post-trains custom models and agents on each enterprise's own data using reinforcement learning
- Forward-deployed engineers and researchers build evals, environments and production agents for customers
- AC2 Agent Cloud trains, serves and improves open-weight models in production
Why it matters
- Customers own the models, which reduces dependence on general-purpose frontier APIs
- Customers include DoorDash, Cognition, Mercor, Microsoft, NVIDIA and Harvey, per the company
- Reported annualized revenue near $50M about a year after its 2025 launch
Best fit
- Enterprises with proprietary workflow data and agent tasks that frontier models handle poorly
- AI-native product companies wanting cheaper, faster custom models built on open weights
- ML teams that want managed RL post-training infrastructure using their own harness
Watch-outs
- Founded in 2025 with a small team and a short enterprise track record
- AC2 self-serve platform is still in private beta, with limited public documentation
- Engagements depend heavily on forward-deployed staff, which may limit capacity and scale
Unlock Full Intelligence
Pro members get the complete briefing
- Complete analyst verdict
- Risk assessment
- All 6 Bluescore dimension scores
- Traction signals
- Bluescore history & momentum tracking
- Unlimited company access
Applied Compute is one of the most credible specialists in enterprise RL post-training. It has blue-chip design partners, reported revenue of about $50M a year and backing from Benchmark, Sequoia, Kleiner Perkins and Lux. Buyers with high-value, data-rich agent workflows should run a structured pilot against a clear baseline, either a frontier API or an in-house fine-tune, and compare quality, cost and latency. Before committing, check delivery capacity, how mature AC2 is, model and IP ownership terms, and data-handling controls.
Signal Provenance
Latest observation date and source for each signal category
| Signal Category | Last Updated | Source |
|---|---|---|
Funding Signals Last funding round, Funding stage | Oct 31, 2026 | Montclare research agents |
Hiring Velocity Headcount, Headcount growth | Never | Montclare research agents |
Technical Output Research papers, GitHub activity | Oct 28, 2026 | Montclare research agents |
News Sentiment News mentions, News sentiment | Sep 30, 2026 | Montclare research agents |
Partnership Signals Enterprise partnerships, Distribution partners | Oct 31, 2026 | Montclare research agents |
Revenue Indicators ARR estimate, Revenue growth | Nov 3, 2026 | Montclare research agents |
Leadership Signals CEO tenure, Founder-led | Oct 18, 2026 | Montclare research agents |
Competitive Position Category rank, Market share | Never | Montclare research agents |
Why this Bluescore?
6 dimensions, each scored 0–10 against all — companies. Each adds its weight × score to the Bluescore — see methodology