Agentic AI for bio-chemistry data analysis

The adaptation layer between
modern AI and real lab data.

LumiGen is an agentic AI system that discovers the right analysis tool, configures it for your specific experiment, and ships a working, verifiable pipeline — automatically. Scientists shouldn't need ML expertise to use ML tools.

See how it works → Request early access
The Company

We build the infrastructure that bridges modern AI to real bio-chemistry data analysis.

Modern AI lives in chat. Real bio-chem analysis lives in fragmented, custom tooling. Today, scientists bend their experiments to whatever tool they can configure. LumiGen flips that: an agentic system that adapts tools to the scientist's data, organism, and protocol — and stays reliable when they use it.

01 / MISSION

Make every tool usable

We're not another analysis tool. We're the adaptation layer that makes the powerful tools the field already trusts work on every lab's data.

02 / APPROACH

Customized · Verifiable · Reusable

An agent discovers, configures, validates against the scientist's own criteria, and ships a pipeline they can inspect, correct, and re-run.

03 / FOUNDATION

Built on published proof

The two halves — agentic algorithm discovery and microscopy automation — already exist as peer-reviewed work from our founding team. We're productizing research, not speculating.

The Problem

Bio-chem data analysis drags scientists down.

Across bio-chem labs the same pattern repeats: roughly 60% of analysis time is lost to tool-wrangling instead of science. It breaks in three places.

01

Finding the right tool is its own project

Dozens of narrow options. Commercial alternatives cost $10K+ and run years behind the research frontier.

02

The right tool breaks on your data

A brain-tissue model fails on cardiac organoids. Every lab needs custom ML — almost none have an ML researcher.

03

When no tool fits, scientists become the tool

Hand-scoring frames. Tuning by trial and error. PhD time spent like RA time.

What biologists actually need

Customized tools that fit their science — and stay reliable when they use them.

CustomizedVerifiableReusable
The Solution

An agentic system that adapts tools to scientists.

A scientist + agent loop that discovers candidate tools, configures them, and validates against the scientist's own success criteria — until the goal is met.

1
Scientist

Upload sample data

A representative slice — a few raw images or frames — enough to ground the agent's plan in real data.

2
Scientist

Provide criteria & annotations

A few annotated examples or success criteria — what counts as a correctly segmented cell. The agent's "north star".

3
Agent loop

Self-evolving build & validate

Discover candidate tools, configure, run on the sample, evaluate against criteria, refine — automatically, until it's right.

4
Scientist

Inspect & iterate

Review the result. Meets spec → the pipeline ships. If not, refine criteria and the loop returns to step 2.

The Team

The people building LumiGen.

Muchen Li
Co-founder

Muchen Li

UBC
PhD · UBC — CS, CV & ML

Expertise in multi-agent research systems, computer vision, machine learning, and AI data-analysis pipelines for scientists.

Jiaqi Wang
Co-founder

Jiaqi Wang

McGill
PhD · McGill — Neuroscience

Expertise in neuroscience, daily imaging/video analysis, and grounding product decisions in real scientific workflows.

Chixiang Lu
Co-founder

Chixiang Lu

HKU
PhD · HKU — CS + Biochemistry

Expertise in AI-microscopy, computational methods, and wet-lab biochemistry, with 2× Nature papers in the field.

Let's talk.

Scientists shouldn't need ML expertise to use ML tools. We make every analysis tool work on every lab's data — then we make that knowledge open.

Connect with us