AI / Data Engineer (Founding) at MUFG Global Service Private Limited (2024-11 – Present)
Sole architect. Designed, built, and shipped MUFG's internal AI platform end-to-end from scratch.
- AI Marketplace (0 → org-wide adoption): Independently designed and built MUFG's internal AI Marketplace - containerised, pluggable AI agents (RAG, NL2SQL, classifiers, summarisers) with metadata-driven routing and dynamic agent selection. Drove adoption across 15+ business teams by working directly with each team to map their use case onto the platform.
- Executive Engagement & Standards: Presented platform architecture and engineering roadmap regularly to senior MUFG leadership; defined engineering standards adopted across the AI team.
- Conversational Data Catalog + NL2SQL: Designed a metadata-aware NL2SQL system serving 30,000+ enterprise dashboards at hundreds of queries/day, letting business users self-serve data questions in plain English. Enforced credential-propagated two-tier security so adoption didn't compromise governance - a Collibra-grade backbone built in-house, avoiding a costly commercial license.
- DB2 → Snowflake Migration Platform: Architected a migration suite spanning 5 data warehouses and 100,000+ columns; cut migration timelines by 40%, directly reducing project cost and time-to-value for stakeholders.
- Pluggable RAG Orchestration Framework: Built a runtime-swappable RAG framework so new enterprise data sources could be onboarded with zero code changes - reducing the lead time between a business team's request and a working AI agent.
ML Research Intern at India–Taiwan Collaborative Lab (2024-07 – 2024-09)
- Designed CT lesion segmentation pipelines achieving 0.6 Dice score; improved accuracy 15% through systematic CNN architecture benchmarking.
iOS Developer Intern at Infosys (2024-04 – 2024-05)
- Delivered a Hospital Management System in 3 weeks; agile practices reduced delivery time by 25% and improved operational efficiency by 40%.
Machine Learning Intern at CSIR (2024-02 – 2024-06)
- Built UNet and YOLOv8 robotic-vision pipelines improving arm accuracy by 18%; CNN benchmarking drove 22% performance gains.
- Research contributed directly to two peer-reviewed, IEEE-indexed publications.
Internship - ML · Predictive Modelling at Apollo Tyres
Tyre Burst Prediction
- Built ML models preventing 30% of tyre burst incidents and reducing quality testing costs by 20%.
