I ship complete applications end-to-end — from a custom UI design system to an LLM reasoning engine to an ACID-backed database — without waiting on a second pair of hands.
Full-Stack Engineer who ships complete applications — frontend, backend, and database — solo. Built a production-style SaaS job tracker on Next.js/FastAPI and an AI compliance-verification platform with LLM-based reasoning, deployed on AWS. Grounded in system design, microservices, and cloud fundamentals. Sole backend developer, HCL Tech Hackathon 2026. Seeking a Full-Stack role to build production-grade web and mobile-ready apps.
001 — Build
Independently shipped a local-first SaaS job tracker end-to-end — custom design system, Next.js/TypeScript frontend, LLM matching engine, 100% client-side data via IndexedDB.
002 — Architect
Architected a 3-gate verification cascade — exact match → semantic embedding → LLM reasoning — resolving 59% of 426 elements without an LLM call and cutting inference cost by 50%+.
003 — Hardening
Delivered a banking backend with ACID-compliant transactions, JWT auth across 3 roles, and zero data-consistency failures under concurrent load — solo backend dev on a 3-person team.
Delivered a production-grade banking system's core transaction infrastructure with zero data-consistency failures under concurrent load, achieved by designing a modular FastAPI / SQLite architecture with ACID-compliant transaction handling and JWT-based authentication across three user roles.
Reduced booking and balance-conflict risk to zero across customer, admin, and staff roles — measured by full concurrency-safeguard coverage — by implementing role-based access control fully integrated with the frontend UI.
Delivered the complete system design under a hackathon deadline — evaluated by industry judges on code quality and production readiness — by following Agile sprint cycles as the sole backend developer on a 3-person cross-functional team.
Independently designed and shipped a full-stack, local-first SaaS job tracker end-to-end — custom UI design system, Next.js / TypeScript frontend, and an LLM-powered matching engine that automatically scores and ranks job descriptions against a user's resume, keeping 100% of data client-side via IndexedDB for complete privacy.
Automated job-listing extraction using Cheerio web scraping with deep-link resolution and duplicate-detection logic, eliminating manual data entry across the application.
Case on GitHub ↗Architected an end-to-end platform using FastAPI, Playwright, and LangChain to crawl websites and verify compliance against policy documents — discovering 101 routes and extracting 695 UI elements for automated review — deployed on AWS EC2 with S3-backed document storage.
Engineered a three-gate verification cascade — exact match → semantic embedding similarity → LLM reasoning — resolving 59% of 426 elements without an LLM call, cutting inference costs by over half while generating JSON/HTML/CSV audit reports validated with RAGAS evaluation.
Case on GitHub ↗Engineered a deterministic clinical data processing engine using the Medallion Architecture to transform raw EHR, vitals, and lab CSV feeds into validated Gold Layer assets — processing 500+ patient records with automated risk scoring, anomaly detection, and YAML-driven threshold alerting.
Optimized backend query performance by 35% using SQL JOINs, indexing, and stored procedures within a modular FastAPI backend with Redis caching. Clinicians can modify anomaly thresholds in rules.yaml without touching a single line of Python — and can generate an on-the-fly, AI-written risk summary for any patient straight from the dashboard.
Case on GitHub ↗Built an enterprise-ready dynamic evaluation engine for job interview preparation — featuring adaptive question memory that mathematically prevents duplicate questions, an anti-cheat timer using browser API hooks via useRef, and real-time analytics across Quantitative, Logic, and Verbal categories.
Deployed across a decoupled cloud architecture — Vercel (React frontend, CI/CD), Render (Spring Boot Docker container), and Supabase (PostgreSQL Session Pooler) — with a custom Python auto-seeder that bulk-injects questions via REST.
Case on GitHub ↗Built an AI-powered, deterministic IT support engine that cross-references incoming tickets against a verified historical knowledge base using TF-IDF vectorization and cosine similarity — then delegates final resolution synthesis to Groq LLaMA 3.1, grounded purely in past company tickets.
Not a generic chatbot — it's a RAG comparison engine that forces the LLM to ground every resolution in actual past tickets. Includes JWT authentication, SQLAlchemy ORM, and CSV/Plotly reporting.
Case on GitHub ↗Full-stack MVP that translates raw engineering metrics into actionable narratives. While traditional dashboards show "Bug Rate: 12%", DevPulse tells you: "This is a test gap — add integration tests to the auth module next sprint."
Layers a rules-based insight engine over raw data — evaluating metrics against review wait times, hotfix counts, and deployment frequency to classify patterns like "Healthy Flow" or "Quality Watch" with one concrete next action per developer.
Case on GitHub ↗Education
Certifications — 7th Semester
Selected after clearing technical assessment rounds and structured industry training. Project evaluated by industry judges on code quality and system design.
Delivered an end-to-end web application under extreme time pressure as part of a cross-functional team — rapid shipping and agile collaboration under the clock.