Full-Stack & Backend Engineer
I build reliable backend systems — and I'm now bringing that same engineering discipline to AI & RAG tools for e-commerce.
01 — About
I'm Rajdeep, a backend-leaning full-stack engineer based in Kolkata, India, with 4+ years building production systems for B2C and B2B products — REST APIs, event-driven architecture, databases at scale, and secure auth systems serving real users.
I'm now specializing in RAG and agentic AI, applying the same rigor around reliability, data integrity, and performance to AI-powered tools for e-commerce brands — product Q&A assistants, support automation, and personalized shopping experiences.
I'd rather ship something boring and reliable than impressive and fragile.
02 — Current Engagement
I'm currently embedded on the Customer Portal team at eSentire, an enterprise cybersecurity SaaS platform. Over the past 12 months:
Migrated 11 report-viewer chart/table renderers off raw D3/Bootstrap onto a shared design system, then resolved roughly a dozen export and data-correctness defects across UI and API.
Traced a site-name resolution bug across four backend services, replacing a brittle dependency with a direct database lookup — fixing broken reports for multi-site accounts.
Built bulk download/delete end-to-end — a single presigned-URL endpoint with per-file permission checks, partial-failure handling, and configurable batch limits — replacing one-at-a-time downloads.
Replaced a manual deprovisioning process with an automated termination workflow — a new microservice endpoint, orchestration logic, and UI guardrails.
03 — What I Do
Product Q&A chatbots, support automation, and retrieval-augmented tools built on your own catalog and docs. My current focus area.
APIs, databases, and event-driven systems designed to hold up under real traffic — not just demo well.
Latency optimization, zero-downtime deployments, and the observability to know your system is actually healthy.
04 — Selected Work
Led a 100K+ record migration (MongoDB → PostgreSQL) during a Strapi v3→v4 re-platform. Delivered in 45 days with under 5 minutes of downtime, zero data loss, and zero post-migration bugs — including schema design, idempotent ETL, and a verified rollback plan.
Voice-based, role & level-aware mock interview tool with automated rubric scoring and feedback. ~4.7/5 user satisfaction, ~25% lower LLM runtime cost through caching and batching, accessibility score raised 76 → 96.
A retrieval-augmented Q&A chatbot — my current build as I go deeper into RAG and agentic AI for e-commerce use cases.
05 — Tech Stack
06 — Contact
Based in Kolkata, India — working with teams across the US, UK, EU, Australia, and the Middle East. Open to freelance and contract work.