Abdullah Hanif

Machine Learning Engineer | Bioinformatics

Saarbrücken, Germany

I’m a machine learning engineer with a background in computer science and bioinformatics. I build machine learning systems for biological data from start to finish: data pipelines that are validated and reproducible, models ranging from protein and genomic language models to classic learners, evaluation that holds up to scrutiny, and services and web applications that people can actually use. Before moving into bioinformatics, I spent nearly four years as a software developer, which is why I care as much about whether a system runs reliably in production as I do about the model behind it.

Experience

Master’s Thesis — Genomic Language Model for AMR Prediction

DATABIOMIX (industry partner) & Chair of Clinical Bioinformatics, Saarland University / HIPS · Aug 2026 – Present

Instead of running a lab test to find out which antibiotics a bacterial strain resists, my thesis predicts it directly from the strain’s genome. I’m building a transformer-based genomic language model for this with DATABIOMIX, co-supervised by the company’s CEO and a professor at Saarland University, and trained on curated resistance data benchmarked against real lab results.

Genomic Language Model — pipeline diagram

Research Assistant, Data Science & Public Health

Department of Epidemiology, Cornell University · Nov 2023 – Mar 2024

I built predictive models for infertility risk and disease patterns (tuberculosis, hepatitis, polio) from public health data, to help identify where healthcare resources were needed most.

Backend Developer & Data Analyst

AseelApp, Arlington, VA, USA · Sep 2021 – Jul 2023

I built backend systems and REST APIs in Python/Django for an e-commerce platform and helped migrate it from a multi-page to a single-page architecture. I also designed its database and turned its SQL data into sales and product insights with Redash.

Software Developer (part-time)

Rahanet ISP, Herat, Afghanistan · Oct 2019 – Jul 2021

Alongside my bachelor’s degree, I built StockOCR, an internal inventory tool that uses Tesseract OCR to read serial numbers and MAC addresses straight from equipment photos, together with its check-in/check-out workflow.

More projects

PhageMatch — Predicting phage-bacteria interactions

Project lead · Saarland University · Sep 2024 – Apr 2025

I led a 4-person team building a model that predicts which phages (viruses that infect bacteria) will work against which bacterial strains, for phage therapy as an alternative to antibiotics. We used graph neural networks and self-supervised learning to make the most of limited labelled data, and I built the Snakemake pipeline from raw data to evaluation. We placed 3rd at Startup Weekend Saarbrücken 2024.

PhageMatch — pipeline diagram

GitHub →

Education

M.Sc. Bioinformatics — Saarland University

Oct 2024 – Feb 2027

Machine learning, neural networks, genomic data analysis & NGS, single-cell bioinformatics, epigenetics & epigenomics, microbiome data analysis, computational drug design.

M.Sc. Applied Data Science & Analytics — SRH University Heidelberg

Oct 2023 – Sep 2024 (coursework completed)

B.Sc. Computer Science — Herat University

Jul 2017 – Jun 2021

Skills

Programming

Python, TypeScript, SQL, R, Bash

Machine Learning

PyTorch, Hugging Face Transformers, scikit-learn, LightGBM, Optuna, SHAP, protein language models (ESM-2, SaProt), graph neural networks, self-supervised learning, conformal prediction

MLOps & Deployment

Docker, GitHub Actions, Google Cloud Run, ONNX Runtime, MLflow, Hugging Face Hub, FastAPI, Django, Next.js, React, Git, Linux

Pipelines & Data

Nextflow, Snakemake, Hydra, Pandera, DuckDB, Parquet, pandas, NumPy

Genomics & Omics

RNA-seq (DESeq2, edgeR), single-cell (Scanpy, Seurat), epigenomics (Bismark, MACS2), multi-omics (MOFA), microbial genomics (Prokka, Sourmash), Biopython, Bioconductor

Protein & Structural

ProteinGym, AlphaFold2 structures, Foldseek, ColabFold, HHblits/HHsearch

Contact

I’m open to full-time roles in machine learning and applied ML, especially where models meet biological or health data. I’m also always happy to talk about protein language models, genomic ML or antimicrobial resistance.