Home›Companies›Probe

The Agent-Native Research Stack

Probe is building the agent-native research stack. Our goal is simple: help AI/ML teams run better experiments, faster. We believe research will drive progress and everyone deserves to have their own personalized lab that iterates faster than ever before. Probe brings core research tools, shared context from past experiments, and visibility into each run together in one platform. Helping teams spend less time setting up infrastructure, supervising agents, catching teammates up and more time testing ideas. Built for helping research labs experiment faster while keeping every researcher and agent aligned, and helping AI/ML teams setup a research capability without building a lab’s infrastructure first.
Active Founders
Co-Founder and CPO @ Probe. Previously studied EECS and conducted SpecDec research at Berkeley. Worked on inference for LoRA fine-tuned models at Databricks. Developed machine translation models at Apple that were SOTA with Apple data. Turned down ML at Together AI to start Probe.
Co-Founder and CTO @ Probe. Previously studied EECS at Berkeley. Worked at Willow Voice (P25) as first engineering hire; led 0-to-1 development and overhaul of ML infra. Worked at Google on YouTube Shorts (feature went live with millions) and hand interactions for Android XR. Started a drone repair business in middle school. Turned down SWE at Palantir to do YC.
Co-Founder and CEO @ Probe. Previously studied EECS at Berkeley. Repeat founder who built an EdTech startup backed by the Qatar Foundation that scaled to a large user base, and co-founded the Chemistry Catalyst Fellowship with Chemistry VC. Shipped production ML systems at Amazon, Rippling, and Sony. Built DetectBot, a viral AI image manipulation bot that reached hundreds of thousands on Twitter (X). Turned down an FDSE at Palantir to build Probe.
Company Launches
Probe: The Agent-Native Research Stack
See original launch post

Introducing Probe: The Agent-Native Research Stack.

Our goal is simple: help AI/ML teams run better experiments faster.

https://youtu.be/xOF8YLKm8nA

The Problem

Teams are using agents for research, but their work is disordered.

They lose hours verifying what their agents actually did, managing environments, sourcing and setting up compute, babysitting runs, wiring up evals, and re-explaining company context their agents should already know.

On top of that, they have to spend critical time in meetings to share learnings across one another, and figuring out what to try next.

The Solution

So we rebuilt the research stack from the ground up for agent-driven research.

How it works: Probe brings three things together in one unified platform:

  • Agent-native research tools: From experiment tracking, compute setup, and GPU monitoring, to paper discovery, dataset retrieval, and lit-review. Probe brings every research tool into one platform so agents can hit the ground running.
  • Shared context from past experiments: give researchers and agents access to what team has already tried, so they can build on it instead of repeating the same failures and work.
  • Visibility into ongoing research: see what researchers and their agents are working on, track experiments, and steer research without waiting for another status meeting.

We help teams waste less compute on failed runs or tedious setup, and spend less time coordinating work.

With Probe, researchers have the tools, context, and control to take ideas through rigorous experiments and faster iterations, creating more research progress with the same compute.

Our asks

  • If you’re a research lab, please reach out. We would love to help you scale your research team without sacrificing speed.
  • If you’re an AI/ML team, please reach out. We would love to help you stand up a research capability with the infrastructure of a frontier lab.

See Probe for yourself: https://www.useprobe.ai/

Probe
Founded:2026
Batch:Spring 2026
Status:
Active
Location:San Francisco
Primary Partner:David Lieb