Available for backend and AI product roles
I build dependable AI systems after the prototype stage.
Backend AI Engineer focused on production-grade backend systems, LLM-powered workflows, agentic applications, and reliability-focused infrastructure.
I work at the intersection of backend engineering, AI product development, and systems that need to become dependable production software.
My experience includes AI SDR/CRM automation, lead enrichment, campaign orchestration, billing workflows, real-time communication, document extraction, and RAG-based research systems.
Proof
Selected Projects
Terraform-managed GCP deployment for distributed LLM inference with a public FastAPI gateway, private workers, Docker Compose, health checks, and autoscaling-ready infrastructure
Read moreContext7 clone for arXiv papers using LangGraph, scoped and hybrid retrieval, LangSmith tracing, and eval workflows
Read moreNext.js agentic frontend with WebSocket streaming, tool-call rendering, trace timelines, and context diff inspection
Read moreAI hiring assistant that extracts structured resume facts, answers candidate questions and grounding
Read moreIn the wild
Open Source Contributions
Electron hotkeys · Dictation UX
Shipped merged fixes for shared dictation controls, reliable audio cues, and safe Electron hotkey rebinding when an old accelerator is invalid.
Bun migration · CI
Migrated project setup from pnpm to Bun across GitHub Actions, scripts, docs, Playwright, TypeScript, Netlify, Cloudflare, and iOS CI.
JavaScript · Python · Rust
Changed authentication boundaries across three SDKs so cloud clients require API keys while local and self-hosted clients can initialize without one.
Java API design
Replaced a misleading SSH config directory API with an explicit file-path property while preserving runtime compatibility through one fallback.
History
Work Experience
- Reworked the PostgreSQL/PgBouncer connection path so the backend no longer depended on one PgBouncer VM, using multiple endpoints, separate pools, endpoint rotation, circuit breakers, and bounded retries
- Moved VM-hosted Node workers from manual screen sessions to systemd, adding crash restarts, reboot survival, standard status/log tooling, and blue/green deploys where the new version starts before the old one drains
- Fixed Stripe checkout amount conversion for zero-decimal currencies such as JPY, KRW, and VND by applying currency-specific integer minor units instead of treating every currency like USD cents
- Made production crashes easier to debug by moving service logs to journald, adding retention limits to protect disk space, and enabling Node diagnostic reports for fatal errors, uncaught exceptions, and shutdown signals
- Built backend workflows for an AI SDR/CRM product: campaign setup, ICP filters, mailbox/LinkedIn account connection, lead research, draft approval, scheduling, and automated sending
- Designed the campaign orchestration layer that turned user inputs into outreach plans, generated drafts, activation jobs, send schedules, and daily execution tasks across email and LinkedIn
- Built LLM personalization flows using reusable templates, research instructions, per-lead context, and fixed sequence structure, so generated messages stayed consistent while still referencing each prospect's company and role
- Shipped lead sourcing and enrichment workflows that combined CSV imports, scraped websites, LinkedIn data, email discovery, ICP filters, and Typesense fuzzy search to find, rank, enrich, and save qualified prospects into campaigns
- Engineered a sophisticated lead generation tool using Chrome extension architecture and agentic AI workflows
- Architected a custom multi-agent AI system achieving 62% performance improvement: reduced response time from 60s to 23s compared to GPT-research-tool
- Built internal dashboards for tracking client leads and sales
- Led frontend work for RateMyStation, an anonymous newsroom review platform, building review feeds, news director search, job-detail cards, category navigation, subscription flows, responsive layouts, and cross-browser fixes
- Built a Next.js 13 suit-customization frontend and connected product options, customization state, cart, and checkout flows to Shopify
Training
Education
SKNCOE
August 2019 - 2022
CGPA: 9.1
Govt. Polytechnic Malvan
June 2016 - May 2019
CGPA: 8.52
Contact
Connect
Reach out for AI engineering, backend platform, or full-stack roles with strong backend ownership.