thampasaurus REX

Built AI-driven personalisation for Best Next Action (BNA), developing features and propensity models supporting a £45M annual uplift investment case. Productionised BNA scoring on Databricks, serving daily AI recommendations to 2M customers across 2 markets.

Led a team of three developer-researchers to build an LLM model platform and associated products. Integrated Facebook's LLAMA and optimized it using llama.cpp, achieving up to 40 requests per second processing.

Led a team of two developers to design and implement a platform for assessing and mitigating risks associated with machine learning models. Added A/B testing evaluation functionality and enhanced platform capabilities with BI analytics.

Conducted around 15 A/B tests in 'Pyaterochka', 'Perekrestok', and 'Karusel' stores to evaluate the impact of business initiatives. Implemented an MVP for optimal discount selection in the 'Promo Pyaterochka' product based on historical data.

Improved microservices performance by 67% by segregating computational blocks of object recognition models into separate services. Configured Grafana and CI/CD system and increased video surveillance data retrieval speed by 40% through FFMPEG keyframe configuration.

Grade: 4.5/5

Pursued a degree in Control Systems and Robotics at ITMO University, achieving a GPA of 4.5 out of 5.