Computational science · biotechnology · pharmaceutical R&D

Turning biological data into something you can act on.

Computational biology, pharmacogenomics, molecular modelling and machine learning — plus the software engineering that decides whether an analysis can be re-run next year by someone who did not write it. I build things that work for other people, not just notebooks that worked once.

Industry CV See working tools Email me

Open to computational biology, bioinformatics and scientific software positions

Core capabilities

Four things I can do now.

Genomic analysis

Variant annotation, protein-level consequence, population genetics and association statistics against live public resources — Ensembl, VEP, gnomAD.

6 published analyses

Pharmacogenomics

Gene–drug evidence, star alleles, CPIC function categories, and a clear separation between annotation and clinical guidance.

8 pharmacogenes · 3 live tools

Molecular work

Descriptors, drug-likeness filters and docking, backed by three years of actual synthesis and characterisation.

BSc Medicinal Chemistry

Software engineering

PostgreSQL schema design, sandboxed code execution, automated testing and continuous deployment, in production.

2 live systems
The part most graduates can't show

Systems that other people use.

Analysis skill is common. What is less common in a graduate is having had to keep software running when the people depending on it are not you.

  • A hostile trust boundary. ForgeLab grades code written by the person being graded, so every claim the browser makes is re-established server-side, under a permission model with no filesystem writes, no child processes and a hard runtime limit.
  • Shared architecture at scale. Gradify runs 30+ sites from a single kit. Sites consume it and never fork it, which is the only reason a fix can land everywhere at once.
  • A real database. PostgreSQL with version-controlled migrations and row-level security, not a pile of CSVs.
  • Tests that run. Including a self-test for the grader itself, because a grading system that silently breaks is worse than one that is down.
ForgeLab architecture

Honest gaps

Worth knowing before an interview rather than during one.

No industry experience yetMy production software is my own, not a company's Workflow engines are proposed, not usedNextflow and Snakemake are a direction, not a claim No deep learning in productionClassical ML only, so far No GxP or regulated environment experienceFamiliar with the reproducibility principles, not the compliance framework
Where I'd fit

By role.

If you are hiring for one of these, these are the pieces of work most worth your time.

How I work

Three commitments.

01

I will tell you what the data can't support

Including when that is inconvenient for a deadline. An analysis that overstates its confidence costs far more later than one that is delivered with its caveats attached.

02

I build it so it can be re-run

Pinned environments, fixed seeds, documented provenance. The test is whether a colleague can reproduce the numbers without me in the room.

03

I can talk to the chemists

Having done synthesis and characterisation, I can explain a computational result to people who will have to act on it, and take their scepticism seriously.