Computational biologist interested in computational neuroscience and computational biology, exploring how mathematics, computation, and biological systems intersect to tackle modern scientific problems. I build reproducible tools and workflows that turn complex data and simulations into structured, interpretable, and trustworthy results. I also play guitar — and my favourite project is the one where those two worlds collide.
name: Hossam
role: Computational Biologist · Simulation & Scientific Tooling
works on: GROMACS molecular dynamics · tumor growth modeling · reproducible pipelines
curious: computational neuroscience · structural biophysics
also: Flutter apps for guitarists 🎸
principle: "If it isn't reproducible, it isn't a result"- 🔬 Building mdforge — one command from raw GROMACS output to a full, reproducible analysis
- 🎸 Shipping exercio — an offline-first guitar practice app, currently in public beta
- 🧠 Drawn to computational neuroscience — modeling the brain is the problem I most want to work on next
- 🔭 Also interested in structural biophysics, biological modeling, and open reproducible science
exercio — offline-first guitar practice, routine building & progress tracking
Most practice sessions have no structure. Exercio gives them sections with targets, exercises with tempos, and an honest record of what actually happened.
The project I'm proudest of — my two obsessions, guitar and rigorous engineering, in one Android app.
What it does
- 🧬 Guitar DNA — an 8-question profile generates a personalized routine through a fully deterministic, on-device algorithm. Same answers, same routine, every time. No randomness, no network calls.
- ⏱️ Practice mode — section and per-exercise focus timers. Overrunning a target is recorded, not blocked — the app learns from what you really did.
- 🥁 Metronome — 6 synthesized sound packs, hold-to-sweep tempo, running as an Android foreground service so it survives a locked screen.
- 📈 Progress tracking — per-attempt tempo, quality, fatigue and tuning logs, with clean-streak and tempo-progress analytics that recompute retroactively.
- 📚 147 practice topics across 15 categories — all stored as data, never hard-coded screens.
- 🎨 6 themes — a token-based design system where a theme changes colors, typography, animation and the metronome's voice without touching feature code.
Engineering worth a look
- Layered pub workspace:
practice_coreis pure Dart with zero external dependencies — no Flutter, no database, no platform code — with the dependency rule enforced by a CI-run architecture checker. - Privacy by construction: no accounts, no analytics SDKs, no crash reporting, no uploads. Everything stays on the phone.
- Native Kotlin audio engine behind a Dart API contract.
"Turns a GROMACS simulation directory into a complete, publication-quality, fully reproducible analysis — with minimal input."
Auto-detects system composition, picks the right analyses, and runs 24 built-in analyses — RMSD, Rg, SASA, H-bonds, RMSF, DSSP, PCA, clustering, salt bridges, contact maps, native contacts, interface and binding-pocket analysis. Every run writes a manifest recording library versions, git commit, input fingerprints, parameters and runtimes. Plugin architecture for custom analyses; config-driven for batch/HPC; containerized for a reproducible environment.
Python 3.10+ MDAnalysis MDTraj GROMACS Docker
A publication-quality pipeline for a 100 ns CHARMM36 simulation of α-zein (UniProt A8HNE1, 187 residues) built from an AlphaFold 3 model. 20+ self-contained analyses — RMSD with moving-average plateau detection, RMSF, Rg, SASA, H-bond networks, PCA + free-energy landscapes, DCCM, DSSP, native contacts, clustering, residue interaction networks. Trajectories are preprocessed once into a lean protein-only file for fast downstream runs; outputs are 300 DPI PNG + PDF figures, CSV datasets and an auto-generated interpretive report.
Python MDAnalysis MDTraj panedr AlphaFold 3
Five progressive notebooks modeling tumor growth as coupled diffusion, logistic proliferation and angiogenesis. Solves the Fisher–Kolmogorov PDE by finite differences with a 5-point Laplacian stencil on grids up to 200×200, sweeping diffusion coefficients, proliferation rates and vessel capacity — ending in a fully coupled tumor–vasculature field where vessels adapt to nutrient demand.
Python NumPy Matplotlib Jupyter
"Predict the paper → forge the study guide → drill it to retention."
Turns lecture slides, past papers, lab manuals and professor transcripts (English or Arabic) into source-grounded study guides, confidence-tagged exam predictions and spaced-repetition drills. Evidence is explicitly weighted — past papers strongest, textbook mentions weakest — and AI-inferred predictions are visibly separated from what a professor actually said. Ships as both a Claude chat skill and a Claude Code kit with persistent progress tracking.
Python Claude Skills Spaced Repetition
Languages
Scientific Python
Molecular Dynamics & Structural Biology
Mobile & Tooling
Between simulation runs, I'm usually on the guitar — which is exactly how exercio got written. A good solo and a good pipeline are built the same way: patient iteration until it's right.
A brutally honest look at my focus levels between experiments 👇
I'm a firm believer in free and open source software — it's the reason I adore GitHub and platforms like it. Knowledge that can be read, copied, corrected and built upon is knowledge that survives. That belief isn't decorative: every public repository I own is MIT licensed, and exercio is free software down to its fonts.
Science and free software want the same thing. A result you can't reproduce isn't a result; a program you can't read isn't knowledge. Both only work when people are allowed to look inside.
— Richard Stallman
"If it isn't reproducible, it isn't a result."
