Mihretab Nigussie

I've built ML-backed products and data platforms across healthcare, personalization, and adaptive learning, taking work from problem framing and experimentation through deployment and monitoring.

At HakimHub, I led development of a clinical insights and matching engine that aligned 20K+ patient records to treatment guidelines and improved physicians’ diagnostic accuracy by 18%. I built Python pipelines for cleaning, validation, feature generation, and reproducible scoring, partnering with clinicians to translate guideline logic into measurable decision rubrics.

At ReteEat, I deployed collaborative-filtering recommendations for 50K monthly diners, delivering a 12% CTR lift over a popularity baseline. I also orchestrated Kafka, Spark, and PostgreSQL streaming ETL processing 5M events per day with under two-second latency.

I'm also building a production-style support copilot using RAG, agent feedback, outcomes, and evaluation loops to improve response quality. I'm particularly interested in LLM data work, supervised fine-tuning, alignment-style evaluation, and RL approaches.