I'm a molecular biologist and entrepreneur in Basel. I'm working on agentic lab automation and drug discovery. I make open-source instrument drivers and bioinformatics tools that I use in my work and my own lab. I also help build peptide drug discovery pipelines for myself and SP2TX. I am self-taught, and enjoy building things. I use AI to write the majority of my code, but I remain in the loop for major architectural decisions. See the AI code quality disclosure below.

• Building Lambda Biolab, a BSL-1 shared lab and open tooling for bioengineering, since 2022.
• Reverse-engineering lab instruments into open Python APIs, from PCR machines to plate readers to electronic pipettes.
• Running local LLM inference and agent tooling for wet-lab workflows.

Affordable, low-cost, distributed computing in genomics and structural biology

• deepvariant-linux-arm64: Google's DeepVariant, rebuilt to run on cheap Arm64 processors.
• vastai-gpu-runner: Scalable structural biology batch GPU orchestration on Vast.ai with robust R2 storage and crash recovery.

Instrument control and lab automation

• bentolab: reverse-engineered BLE/Wi-Fi protocol for Bento Lab PCR workstations, with async client, FastAPI wrapper, and CLI.
• wallac-victor2-api: REST/JSON API for a PerkinElmer Victor2 multimode plate reader, driven from Linux.
• dpette-usb-driver: open USB driver for DLAB dPette electronic pipettes.
• i3mega-pipettebot: an Anycubic i3 Mega turned into a sub-$150 disposable-tip pipetting robot (collaboration with qte77).

Structure prediction

• biolab-runners: modular runners for Boltz-2 structure prediction and OpenMM MD.

Judge my code and contributions by their quality, not whether they are AI-generated or whether I have a computer science degree. I have contributed code that was validated and starred by Google devs, and I am paid to ship software that my clients rely on. Consequently, I always take the time to validate the behavior of my code as best I can, because I hate low-effort AI slop as much as anyone else. Every repository has quality gates designed to catch most obvious defects, unsafe changes, and shallow implementations by AI agents. For instance:

• ruff and pyright with tests treated as first-class code
• pytest with coverage, plus branch coverage on changed files
• cognitive-complexity limits and mutation testing on changed code
• bandit, secret scanning, actionlint, markdownlint
• actions pinned to commit SHAs

• Lab and org: Lambda-Biolab
• Consulting: lambconsulting.bio
• LinkedIn: in/antoniolamb
• Email: contact@lambconsulting.bio