Skip to content
View Aderfi's full-sized avatar
  • Spain

Block or report Aderfi

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
Aderfi/README.md

Adrim Hamed Outmani

Pharmacist · ML engineer for drug discovery & pharmacogenomics

I build deep-learning models that predict how genes and drugs interact — bridging pharmacology and computational science to make therapy more personalized.

PharmD (Universidad de Granada) · MSc Bioinformatics (VIU) · 📍 Spain


🔬 What I'm building

Pharmagen — a pharmacogenomic prediction engine Predicting the phenotypic effect of a drug (toxicity, efficacy, pharmacokinetic alterations) from genetic variants.

  • Model: hybrid neural network — Factorization Machines + Transformers with attention — trained on ClinPGx & dbSNP (86.9% accuracy).
  • Pipeline: NGS variant extraction and standardization in Python (ClinPGx, dbSNP, PubChem, Ensembl VEP).
  • Engineering: hyperparameter search with Optuna, class-imbalance handling via Adaptive Focal Loss.
  • v2 — graph architecture (in progress): a graph-based pharmacogenomics library — molecular graphs (RDKit → PyTorch Geometric) and positional genomic graphs for gene variants, feeding a Twin-Tower GATv2 architecture.
  • v2 — production refactor (in progress): rearchitecting the codebase to industry standards — typed data models (Pydantic), a FastAPI service layer, and reproducible ML workflows — hardening a research prototype into maintainable, deployable software.

Durin — clinical software for polypharmacy safety (early development) Tooling to support the care of polymedicated geriatric patients — ingesting a medication list and clinical profile to surface the drug–drug interactions, dosing concerns and cumulative side-effect burden worth a clinician's attention.

  • Foundation built: typed domain models (Pydantic v2) for drugs & patients, an ATC/DDD catalog via a custom scraper, and i18n locale catalogs (ES/EN).
  • Service layer: FastAPI scaffolding in place.
  • On the roadmap: the interaction/risk engine and an AI layer to prioritize and explain risks.
  • Stack: Python 3.14 · Pydantic v2 · FastAPI · BeautifulSoup · Polars

Also exploring: computational rheology of topical pharmaceutical forms (non-Newtonian fluid simulation phi-flow) and a Quarto research portfolio.


🧠 Focus areas

Pharmacogenomics · AI-driven drug design · Graph Neural Networks Deep Learning for molecules · Precision medicine · NGS pipelines


🛠️ Stack

ML / DL: PyTorch · PyTorch Geometric · Optuna · scikit-learn Cheminformatics / Bio: RDKit · NGS/VCF · Ensembl VEP · dbSNP · PubChem Data: pandas · Polars · NumPy Tooling: Python · C++ · Git · Linux (Debian) · Quarto


📄 Selected work

  • TFMAI & Personalized Medicine: pharmacogenomic prediction pipelines and deep-learning models (VIU, 2026).
  • TFGPharmacogenomics of hypertension (Universidad de Granada, 2022).

📫 Contact

LinkedIn · Email

PharmD who taught himself to code — I like turning biological questions into models. Open to research collaborations and PhD/CDT opportunities in AI for drug discovery.

Pinned Loading

  1. Pharmagen Pharmagen Public

    Design of a model to predict the effectiveness of a pharmacological treatment.

    Python 2