Business Technology Solutions Associate in ZS's AI Centre of Excellence, Pune.

Developed an end-to-end RAG pipeline for a natural-language-to-SQL analytics agent, combining semantic retrieval over schema context with LLM query generation and ChromaDB-based query caching to cut latency and token cost. Built a label printing desktop application in Python and PostgreSQL deployed across all Instamart warehouses pan-India, saving approximately Rs. 26 lakh annually. Engineered a face recognition attendance system using InsightFace, FAISS and Flask, processing around 1,500 check-ins daily with offline resilience during network outages. Developed the daily picking algorithm with expiry-ordered batch allocation and layout-aware sequencing, reducing picker travel time across the warehouse network.

Co-founded a software services company, owning the full client lifecycle from pitch and requirement gathering through scoping, development and delivery. Delivered service-based ERP systems and loan management platforms for clients across industries, managing parallel engagements end to end.

Architected the backend for Fundsmama, a consumer lending platform with 500K+ Play Store downloads and 50K+ monthly active users. Built a multi-lender aggregation engine integrating 23 loan provider services, with Kafka event streams propagating real-time loan status updates. Designed approval state machines for disbursement tracking and repayment reconciliation with RBAC and audit trails, and implemented a loan eligibility engine with Redis caching and rate limiting. Also built an inventory management system for Bombay Tyres and an S3-backed document management system for Rupali Logistics.

Built an enterprise-wide ERP system using ReactJS, Node.js, ExpressJS and MongoDB, deployed on AWS EC2 and supporting 500+ concurrent users, with asset management, project and task tracking, team management and dashboard analytics modules. Implemented JWT authentication with role-based permissions and audit logging, and designed versioned REST APIs with validation middleware, rate limiting and structured error contracts.

Engineered a universal data context layer unifying SQL, NoSQL and flat-file sources into a single knowledge graph, giving AI agents one reliable interface to structured enterprise data. Built metadata tagging and schema lineage tracking, schema-aware query planning to keep generated queries grounded in the real data model, and TTL-based semantic caching. Built Next.js and React dashboards for Gadigoda.

Built client web applications using the MERN stack and PHP.

Built the complete 99Hectors website end to end, including a content management system, lead management with end-to-end lead tracking (capture, assignment, status and follow-up history), an admin dashboard for non-technical users, and media upload and management.