Biomedical engineer turned data scientist turned Machine Learning Engineer. Currently training neural nets for oligonucleotide drug design and delivery at GSK, as well as exploring agentic drug discovery. Outside of work, I really enjoy building all sorts of side projects — from things that help me in my daily life to bigger, more ambitious and visionary projects like agentic systems, novel human-to-agent interfaces and the design around them, and open-source developer tools. I enjoy tinkering, self-hosting, and running a homelab.
luischav.es · LinkedIn · London, UK
AI-driven coding and agents have completely lowered the barriers to entry, letting me fully express my creativity through code. These span agent orchestration, novel interfaces, developer tooling, and more.
-
🧵 mainthread — An agentic harness for creating trees of agents that communicate for long-running tasks and autonomous research. I previously explored agent-to-agent communication with c2c (Oct/Nov 2025) before it became mainstream — predating features like Claude's agent teams, OpenAI Codex's multi-agent support, and Google's Agent-to-Agent protocol. That early thinking was folded into mainthread.
-
🖥️ always-on-claude — Guide for setting up an always-on Claude Code server with Tailscale + Mutagen, accessible from any device.
-
🚇 TfL Uptime — Transport for London as a tech status page.
-
📊 parquet-peek — Browse, sort & filter Parquet files directly in your browser. Available on the Chrome Web Store.
-
🎞️ slidev-marimo — Marimo notebook integration for Slidev presentations (Pyodide/WASM + live kernel).
-
📝 Early packaging & knowledge sharing (2019--2021) — During my MSc and the height of the COVID pandemic, I got into hackathons, self-hosting, and understanding more about computer architectures. This was my earliest work around Python packaging and technical writing, where I posted several articles on Towards Data Science (bar chart races with Plotly, automating COVID-19 reports, interactive COVID-19 dashboards) and published my first open-source packages: raceplotly (45k+ downloads) and nb2medium (17k+ downloads).
-
📖 etymology-for-all — Free etymology for everyone.
More projects not yet posted about: society (agent-to-agent communication powered by mainstream transport protocols like SSH and HTTP), refine (improve agent skills the more you use them), eval-claude (daily evals on Anthropic models to monitor model degradation).
- marimo — 7 merged PRs: dotenv loading, pixi dev environment, Azure OpenAI, multi-column sorting, ACP dynamic model selection, islands fix, docs
- happier — 2 open PRs: CLI permission escalation fix, session decryption parallelization
- kedro — dynamic session loading docs for non-Jupyter environments
- kindtech — docs rewrite + codebase modernization with narwhals
- wigglystuff — TangleSelect component
- Also contributed to: Inspect AI (UK AI Safety Institute), roborev (Wes McKinney), pixi
Active Discord member in: marimo, polars, pixi, happier (say hi!)
DataKind UK — Community Committee Member & Volunteer Data Scientist. Built LASSO models, Streamlit apps, and data pipelines for charities including Learning With Parents, Citizens Advice, and Amos Bursary.
Core expertise is ML, but broad interests across the stack: low-level programming, full-stack apps, iOS apps (colour perception tools for colourblind people), Chrome extensions, data pipelines, data visualisation, open source, novel interfaces, wearable fitness tracking, and data-for-good.
- 2025 -- ML Engineer, GSK — transformer models on RNA for oligo drug design & delivery, multi-node multi-GPU training on Vertex AI, pioneered Claude Code adoption across AIML org
- 2024 -- 2025 Data & ML Engineer, Merqato — demand forecasting for European fresh produce wholesalers (engineer #3)
- 2024 Career break — upskilling
- 2023 -- 2024 Sr. Data Scientist II, Visa Europe — Technical Excellence team, scaled ETL across 27 European banks, led inner-source team of 8
- 2021 -- 2022 Sr. Data Scientist I, Visa UK&I — built VisaVerse, adopted by 400+ data scientists globally
- 2020 -- 2021 Data Scientist, Visa UK&I — led GBP 1M consulting project, GPU-accelerated ML R&D
- 2019 -- 2020 MSc Health Data Science & ML, Imperial College London — Distinction, 3rd/25
- 2015 -- 2019 MEng Biomedical Engineering, Imperial College London — First-Class Honours, iGEM 2018
luischav.es — London, UK