02.
Experience
// from enterprise Java to AI systems — the path, in order
Aug 2025 — Present
Founding Engineer (FullStack + AI)
AI systems engineer
- Architected an agentic RAG pipeline using E2B sandboxes, exposing a 2.3M+ record legal knowledge base as MCP tools for dynamic retrieval and citation with isolated execution per conversation.
- Built and vectorized the legal knowledge base using LlamaIndex embeddings and Weaviate, evolving the platform from plain LLM calls to a fully agentic retrieval pipeline for accurate legal Q&A and contract analysis.
- Integrated Stripe subscription billing with webhook handling for charges, refunds, and cancellations.
- Built Google Drive and OneDrive integrations using GCP and Microsoft Graph API for seamless document import/export.
- Engineered real-time AI response streaming via SSE, reducing latency from 3–5 seconds to under 1 second; implemented session recovery reducing failed generations by 30–40% during interruptions.
- Resolved Prisma connection pool exhaustion under parallel load by tuning pool settings, restoring stability at scale.
- Integrated Langfuse for LLM observability across workflows; developed dynamic contract editing interfaces using React and TipTap.
Mar 2025 — Jun 2025
Freelance Software Engineer (Contract)
full-stack · security-minded
- Contributed to a full-stack platform (Next.js T3, Prisma, PostgreSQL) for clinician–patient workflows.
- Engineered an AES-encrypted note management system to preserve confidentiality of sensitive clinical records.
- Implemented RBAC and strict user-linked data isolation; designed secure password recovery flows while maintaining encryption integrity and access constraints.
- Built responsive interfaces using React, TanStack Query, and ShadCN.
Sep 2023 — Jul 2025
Software Engineer · Kolkata, WB
enterprise Java engineer
- Implemented a microservices architecture using Spring Boot, improving API response time by 25% and reducing server load by 30%.
- Improved performance through asynchronous service communication using Kafka, increasing throughput by 35% and reducing request latency.
- Developed internal monitoring tools with optimized MySQL queries and advanced data aggregation, improving diagnostic efficiency by 40%.