Flipkart GRiD 7.0 National Finalist | AEH Intern @ Accenture | CSE @ IIIT Guwahati '27 | 📍 Guwahati, India | 📱 +91-7542049250

CSE student at IIIT Guwahati. I build backend services in Java/Spring Boot and like working on real-time systems, search, and security research. Currently interning at Accenture (AWS Data & Analytics), building an AI-assisted health insurance claim assistant on AWS serverless. National Finalist (Top 48 of 1.8 lakh+) at Flipkart GRiD 7.0; lead author on PUF-based authentication for UAV swarms.

May – Jul 2026 · AWS Data & Analytics · 8-week internship

Working in the AWS Data & Analytics track on an AI-assisted Health Insurance Claim Assistant that runs fully on AWS serverless services. Users upload documents, ask policy questions, and submit claims through a single chatbot over a WebSocket connection — no forms or portals.

• Built a serverless backend on AWS Lambda + API Gateway (WebSocket), with DynamoDB for chat history and claim records.
• Used Amazon Bedrock as the first stop for every message to detect intent (conversation / upload / submit) and generate a plain-language reply.
• Grounded policy answers with RAG — a Bedrock Knowledge Base over policy PDFs in S3 — so coverage questions are answered from the actual policy.
• Generated pre-signed S3 URLs for secure, direct document uploads, so files never pass through the backend.
• Orchestrated claim processing with AWS Step Functions: Textract extracts document text, then parallel document-verification and eligibility checks run via Bedrock, then a final claim record is created.
• Pushed real-time claim status updates back to users over the API Gateway WebSocket connection.

Tech: AWS Lambda, Amazon Bedrock, API Gateway, S3, DynamoDB, Step Functions, Textract, WebSocket

B.Tech, Computer Science and Engineering
Indian Institute of Information Technology, Guwahati (IIITG) | Aug 2023 – Present
CGPA: 7.33

XII, CBSE — Cambridge School Silao, Nalanda, Bihar | 89% | 2022 – 2023
X, CBSE — DAV Public School PGC, Bihar Sharif, Nalanda | 89% | 2020 – 2021

• DSA, OOP, OS, DBMS, Networks, Compilers, Computer Security
• REST APIs and microservices
• Real-time systems (WebRTC, WebSocket/STOMP)
• AWS serverless (Lambda, Bedrock, API Gateway, Step Functions, DynamoDB, S3)
• Search, ranking, personalization
• Auth and security (JWT, OAuth, RBAC)
• Schema design (3NF/BCNF) and indexing

• LeetCode — Knight, 1968 rating (Top 3%), 1200+ problems solved
• Codeforces — Expert, 1600+ rating
• CodeChef — 1400+ rating
• HackerRank — 5★ in Java, C, C++
• JEE Advanced 2023 — AIR 15,864
• JEE Mains 2023 — 98.11 percentile

Jan – Apr 2026 · C++, OMNeT++ 6.3 · Supervisor: Dr. Rakesh Matam, IIIT Guwahati

Lead author on a paper proposing a lightweight, PUF-based authentication protocol for UAV swarms. The idea: replace heavy PKI handshakes with something fast that also survives a UAV being physically captured — no private key is stored, the "key" is the silicon itself.

Results (10-UAV stadium-surveillance scenario, OMNeT++ 6.3):

• 0.67 ms UAV ↔ Ground Station authentication (SHA3-256)
• 0.48 ms peer ↔ peer authentication (SHA3-256)
• 10–22× faster than RSA-2048 / ECC-256 in the same setting
• 100% success across all 10 UAV–GS and 45 peer pairs
• 430 B total overhead per UAV pair
• BCH(255,131,18) fuzzy extractor handles 3% PUF noise with a ~4× safety margin
• Dual-hash characterization: SHA3-256 (software) vs. SPONGENT-160 (~2,190 GE, hardware-friendly) with FPGA/ASIC projections

📄 Read the paper

Oct – Nov 2025 · Java 21, Spring Boot, React, WebRTC, AWS

Real-time platform for technical interviews and online classrooms. WebRTC video plus a shared Monaco code editor and whiteboard, deployed on AWS.

Results & engineering choices:

• Code editor sends diff patches instead of full file contents (Java Diff Utils) — ~90% smaller payloads vs. naive sync in local runs
• Horizontally scalable WebSocket layer via Redis Pub/Sub — chat and code sync work across multiple backend instances
• Auto-scaling on AWS driven by custom CloudWatch metrics (active rooms, participants, processing load), not just CPU
• JWT auth (HttpOnly cookies, access + refresh), RBAC, BCrypt, AES-256 at rest, email verification via SES
• Multi-level caching with Caffeine + Redis; Resilience4j circuit breakers + token-bucket rate limiting

🌐 Live: fusemeet.live

Aug 2025 · Python/FastAPI, Node, React, Elasticsearch, ChromaDB, Redis, MongoDB

Flipkart GRiD 7.0 finalist project — a search engine that tries to understand intent instead of just matching keywords.

Results (local benchmarks, 10k+ products):

• ~92% intent classification accuracy on offline eval
• <100 ms autosuggest latency
• <300 ms full search latency
• ~85% Redis cache hit rate
• Hybrid autosuggest with custom scoring (exact > prefix > personalized > fuzzy) — handles typos and abbreviations
• Pipeline: intent classification → vector retrieval (ChromaDB) → cross-encoder reranking → diversification

Dec 2024 – Mar 2025 · Spring Boot 3, Spring Security 6, MySQL, JWT

End-to-end e-commerce backend: product catalog, cart, and admin APIs.

• JWT in HttpOnly cookies (24h access tokens), RBAC for USER/ADMIN, BCrypt (cost 12)
• Product CRUD with multi-level categories, image uploads, stock tracking
• Multi-criteria paginated search; cart persists across sessions with stock-aware validation
• 3NF schema with indexes on hot columns; lazy loading to avoid N+1
• Global exception handling via @RestControllerAdvice, Bean Validation on inputs

Nov – Dec 2024 · Spring Boot, React, MySQL

Folder-based notes with view/edit permission sharing.

• JWT + refresh tokens, HttpOnly cookies, BCrypt (12 rounds)
• AES-256 at rest, rate-limited auth endpoints
• XSS, SQL injection, and CSRF protections on state-changing routes

Apr 2025 · Spring Boot

Waste management prototype with IoT bin sensors, route optimization for collection vehicles, and admin/user dashboards.

Mar 2025 · Python, Pygame, python-chess

• Minimax with alpha-beta pruning, configurable depth (1–7)
• Material + piece-square-table evaluation, move ordering for better pruning
• At depth 5: ~10k positions/sec on a laptop
• Plays around ~1500 ELO informally against humans

Feb 2025 · Python, NumPy, Gymnasium

CEM-trained policy network for OpenAI Gymnasium Lunar Lander.

• Small MLP policy (2× 128 hidden units)
• Multiprocessing for parallel rollouts (~4× speedup)
• Population 50, top 20% elite, ~100 generations
• Averages ~280 reward over 100 evaluation episodes after training

Nov – Dec 2024 · Spring Boot

A small Spring Boot CRUD app for pet records — appointments, vaccinations, medical history, owner info. Mostly an exercise in schema design and JPA.

• 🏅 National Finalist (Top 48 of 1.8 lakh+) — Flipkart GRiD 7.0, 2025
• 🏅 LeetCode Knight — 1968 rating, Top 3%, 1200+ problems
• 🏅 Codeforces Expert — 1600+ rating
• 🏅 HackerRank — 5★ in Java, C, C++
• 🏅 HackerRank certs: Problem Solving (Basic), Java (Basic)
• 🎯 JEE Advanced 2023 — AIR 15,864
• 📊 JEE Mains 2023 — 98.11 percentile

2026

• Accenture — AEH Intern (AWS Data & Analytics), AI health insurance claim assistant on AWS serverless
• UAV Swarm Authentication — research paper (PUF-based, OMNeT++)

2025

• FuseMeet — WebRTC + AWS collaboration platform
• IntelliSearch — Flipkart GRiD 7.0 National Finalist
• ShopSphere — full-stack e-commerce backend
• Cross-Entropy Method on Lunar Lander
• Chess AI with alpha-beta pruning

2024

• Secure Note-Taking Application
• Smart Waste Management System
• PetProject (DBMS in Spring Boot)
• Hit Codeforces Expert

• Interning at Accenture (AWS Data & Analytics) on an AI health insurance claim assistant
• Looking for SWE internships/roles for 2026–27
• Going deeper on distributed systems and a bit of computer security
• Exploring LLMs and vector databases beyond IntelliSearch
• Pushing for Candidate Master on Codeforces

• Real-time systems (WebRTC, WebSocket)
• AI/ML applications and search
• E-commerce / payments
• Auth and security-focused services
• Full-stack web apps
• Competitive programming