Senior Data Scientist | University of Cambridge

I build production-grade bioinformatics pipelines, machine learning systems, and cloud-native platforms that turn genomic data into actionable research outcomes. At Cambridge, I lead end-to-end genomic data provisioning for one of the UK's largest national research cohorts — processing 50,000+ samples, managing terabyte-scale secure transfers, and enabling precision recall-by-genotype studies across rare diseases, immunology, neurodegeneration, and inflammatory bowel disease.

• Clinical Bioinformatics — HLA imputation pipelines (SNP2HLA/Beagle) at 50K+ sample scale, APOE genotyping, GWAS data preparation, multi-platform variant search across 7 genotyping arrays, and genomic QC for WGS/WES/array data
• Machine Learning & AI — Fraud detection with CatBoost (value-weighted loss, isotonic calibration), demand forecasting at 600K+ SKU scale (LightGBM/TFT/N-BEATS ensemble with 150+ engineered features), computer vision invoice digitization (YOLOv5), and clinical NLP phenotyping (SciSpacy + ConText)
• Cloud-Native Platform Engineering — Microservices on Kubernetes (v1.28) with Terraform-provisioned AWS infrastructure (EKS, RDS, ElastiCache, KMS), Docker containers across 17+ services, zero-trust NetworkPolicies, and Prometheus monitoring
• Data Governance & Compliance — GDPR Subject Access Request automation, Five Safes Framework implementation, GPG-encrypted secure transfers with SHA-256 verification and full audit trails

Languages: Python, R, Bash/Shell, TypeScript, SQL, AWK, HCL (Terraform)

Bioinformatics: PLINK 1.9/2.0, BCFtools, SNP2HLA, Beagle, CookHLA, SciSpacy, NCI LDlink API

ML/AI: CatBoost, LightGBM, PyTorch, YOLOv5, scikit-learn, spaCy, GPT-4V/LLaVA integration

Frameworks: FastAPI, Pydantic v2, React 18, TanStack Query, Recharts, Plumber (R), Tidyverse

Infrastructure: Kubernetes, Terraform, Docker, GitHub Actions, NGINX, Prometheus, MLflow, Airflow, Redis

Cloud: AWS (EKS, RDS, ElastiCache, KMS, VPC, S3), Azure (AzCopy, Aridhia SDE)

Production-grade 6-microservice clinical research platform with NLP phenotyping (SciSpacy), visual PII auditing (GPT-4V/LLaVA), NHS Trust data harmonization, React dashboard, and full Kubernetes/Terraform infrastructure.

Value-weighted fraud detection pipeline with CatBoost gradient boosting, expected-value regression, isotonic calibration, m-estimate risk encodings, and business-aligned evaluation metrics.

Enterprise ML pipeline for 600K+ SKU demand forecasting using a LightGBM/TFT/N-BEATS ensemble, 150+ engineered features, custom WRMSSE objective, Airflow DAGs, and MLflow tracking.

Computer vision pipeline for automated invoice field extraction using YOLOv5 with 6-class detection, thread-safe model serving via FastAPI, and hot-reload model management.

End-to-end HLA imputation orchestration for 50,000+ biobank samples — 7-step pipeline with PLINK, SNP2HLA, Beagle, AWK hash-map ID conversion, and 8-locus clinical typing report generation.

Precision medicine toolkit for genotype-stratified participant recall with demographic balancing, age-band stratification, and sex-matched control design for clinical studies.

• Processed 50,000+ samples across HLA imputation batches with automated strand-conflict resolution
• Delivered genomic cohorts of 10,732 samples with cryptographic verification
• Managed 777 GB secure cloud data transfers to research environments
• Harmonized clinical data from 4 NHS Trust formats into OMOP CDM
• Led a team of 6 junior data scientists in genomic data processing and secure data handling
• Enabled research across 10+ disease areas including rare diseases, IBD, neurodegeneration, and immunology
• Supported 45 GDPR Subject Access Requests returning clinical genomic data for patient care decisions

Every repository follows consistent engineering standards: structured Python packages with pyproject.toml, comprehensive test suites, Docker containers, CI/CD via GitHub Actions, Makefile automation, and detailed documentation. Legacy shell scripts are preserved alongside modern Python rewrites to demonstrate the evolution from ad-hoc tooling to production-grade software.

No real participant data, credentials, or proprietary information is committed — all repositories use synthetic data and mock responses, demonstrating data sensitivity awareness and testability.

Cambridge, United Kingdom