I built an end-to-end football prediction platform using 26 seasons of match data, with an XGBoost outcome model and SHAP explanations. Its 2023/24 test-season accuracy was 52.5%, and a 2026/27 backtest beat an Elo-only baseline by 3.7 accuracy points.
I deployed the platform’s Dockerized FastAPI service to Google Cloud Run, with Firestore accounts, per-client rate limits and daily token budgets. I also built a grounded RAG assistant where tool calling improved correct answers from 1/11 to 11/11.
My Football Dashboard is an owner-only monitoring console on Cloud Run, bringing together request search, trace waterfalls, error groups and token costs. Earlier, at Goodwork Labs Services Pvt. Ltd., I implemented a Zee5 tax-breakup fee feature that contributed to an 8% increase in global subscribers.
Across my Android work, I contributed to products including Domino’s Indonesia, Zeta Tech’s Card Suite SDK and Mahindra Last Mile Mobility. I’m now focused on applied AI engineering, building and deploying LLM applications with RAG, agent workflows and evaluation.
