# hiloop

> hiloop is a San Francisco startup building infrastructure for automated research - the compute and observability layer that lets AI agents propose, run, and verify their own experiments. Teams hand hiloop a task, a model or agent, and a metric; hiloop runs snapshottable, forkable compute where every run is captured as a single queryable trace, then returns the best verified improvement across training, data, prompts, tools, and systems. Founded in 2026 by Karan Brar and Thomas Boser, it is part of Y Combinator's Summer 2026 batch.

- **Founded:** 2026
- **Headquarters:** San Francisco, United States
- **Founders:** Karan Brar (Co-Founder & CEO), Thomas Boser (Co-Founder & CTO)
- **Team size:** 2
- **Products:** hiloop compute, hiloop run (CLI), hiloop observability, Autoresearch campaigns
- **Notable:** Accepted into Y Combinator's Summer 2026 batch, Founded by engineers with ML infrastructure experience at Reducto and DynamoAI, Shipped snapshottable, forkable compute with secret brokering and per-run SQL-queryable tracing

## Products & services

- **hiloop compute** — Snapshottable, forkable compute that research agents provision themselves through an API. A fork branches the run tree and inherits the parent's filesystem from its snapshot, so an agent can try an idea, abandon it, or restore any snapshot into a fresh sandbox. Secrets are brokered so the real key never lands in the sandbox.
- **hiloop run (CLI)** — Wrap any command - a homegrown eval script or an off-the-shelf RL loop - with `hiloop run` to get the same provisioning and the same traces, no SDK required.
- **hiloop observability** — Every run is recorded as one trace, with LLM, tool, and stdout events fork-tagged, and tokens and estimated cost rolled up per model. Queryable with SQL.
- **Autoresearch campaigns** — Teams provide a task, current agent or model, and evaluation criteria; hiloop runs research across SFT and other post-training methods, continual learning, prompts, tools, harnesses, and systems, then returns the best verified improvement. Runs hosted or in the customer's cloud.

## Achievements

- Accepted into Y Combinator's Summer 2026 batch
- Founded by engineers with ML infrastructure experience at Reducto and DynamoAI
- Shipped snapshottable, forkable compute with secret brokering and per-run SQL-queryable tracing

## Latest updates

- **2026-07** — Live website describing hiloop as infrastructure for automated research: snapshottable, forkable compute with built-in observability, plus the `hiloop run` CLI.
- **2026-06** — Part of Y Combinator's Summer 2026 (S26) batch, operating as a two-person team in San Francisco.

## Links

- Website: https://hiloop.ai
- LinkedIn: http://www.linkedin.com/company/hiloopai
- Twitter/X: https://twitter.com/hiloopai
- GitHub: https://github.com/hiloopai

---

Profile page: https://yespress.io/hiloop-yc-s26
Published by YesPress — https://yespress.io
Last updated: 2026-07-31