Hi 👋, I'm Shikhar Sinha

Full-Stack Developer | AI Engineer | Backend Systems | Competitive Programmer

• 🎓 Pursuing B.E. in Electronics & Communication Engineering at Bangalore Institute of Technology (Expected May 2027) — GPA: 8.69
• 💻 Full-Stack Developer: Building fast, scalable, and responsive web applications using the modern MERN Stack and Next.js.
• 🤖 AI Engineer: Specialized in Generative AI & Agentic AI — building event-driven LLM pipelines, autonomous agents, and vector-based semantic retrieval architectures (RAG) using Python, FastAPI, LangChain, and LangGraph.
• ⚡ Competitive Programmer: Expert on Codeforces and Knight on LeetCode with a deep understanding of Data Structures & Algorithms.
• 🏆 Achievements: Advanced to Round 2 of Meta Hacker Cup 2025, secured Global Rank 1312 in TCS CodeVita, and shortlisted for multiple national hackathons (Amazon, Adobe).
• 💬 Ask me about: Full-Stack Development, System Design, AI Architectures, or DSA optimization.
• 📫 Reach me at: shikharsinha117@gmail.com

Node.js · Express · MongoDB · React

• Engineered a high-concurrency banking system using MongoDB session transactions to synchronize data and inherently prevent race-condition overdrafts.
• Enforced strict request-level idempotency via compound unique indexes, eliminating duplicate transactions under network retries.
• Architected an immutable double-entry ledger with schema-level hooks to guarantee auditability and financial correctness.
• Implemented a state-driven transaction lifecycle featuring bounded retries and decoupled event-driven notification handling.

Next.js · PostgreSQL · pgvector · Inngest · Gemini API

• Architected an event-driven async pipeline (Inngest) to generate end-to-end courses, keeping API responses under 200ms during heavy LLM workloads.
• Engineered a vector-based semantic caching layer (pgvector) with a two-gate hit detection strategy to drastically reduce redundant LLM calls and latency.
• Built a custom RAG pipeline that chunks, embeds, and indexes course data to ground Gemini API generations and minimize hallucinations.
• Implemented an adaptive remediation loop that analyzes low quiz scores and automatically generates targeted practice content.

Node.js · TypeScript · Socket.IO · Redis · Leaflet

• Built a horizontally scalable real-time location tracking engine handling high-frequency telemetry with $O(1)$ spatial grid geofencing.
• Implemented defensive backend patterns including token-bucket rate limiting, velocity-based drift filtering, and graceful Redis-to-memory state failovers to protect the event loop.

Next.js 15 · Gemini API · Vonage · Prisma · PostgreSQL

• Developed a secure telemedicine portal integrating Google Gemini for symptom analysis and Vonage API for WebRTC real-time video consultations.
• Ensured safe booking and payout execution using atomic PostgreSQL transactions and strict role-based access control.

JavaScript · Chrome Extensions API · Google Gemini API

• Created a lightweight browser extension that extracts video transcripts and interfaces with the Gemini API to instantly generate structured, actionable summaries.

🕵️‍♂️ Fun Fact: I approach debugging like solving a mystery novel — each console.log is a clue leading to the solution!