Computer Science undergraduate at BITS Pilani building across applied machine learning, systems, and backend engineering.
I enjoy working on problems where modelling or algorithmic choices translate into measurable system behaviour — from learning-to-rank for e-commerce search to performance-conscious C++ systems and production-oriented AI/backend applications.
A second-stage product-search reranker combining lexical similarity, dense semantic representations, marketplace-specific relevance features, and LambdaMART Learning-to-Rank.
- Built over 601K query-product relevance judgments across 29.8K queries
- Combined TF-IDF, MiniLM embeddings, lexical/attribute features, and numeric/model-token matching
- Designed query-disjoint train/validation/test evaluation to prevent ranking leakage
- Achieved 0.7050 NDCG@10 on 8,956 held-out queries
- Improved NDCG@10 by 8.72% over TF-IDF and 3.06% over MiniLM
- Added controlled feature ablations, segment analysis, query-level error analysis, tests, and reproducible experiment outputs
Tech: Python · SQL · LightGBM · SentenceTransformers · scikit-learn · DuckDB · Pandas · NumPy
A C++17 single-instrument limit-order-book engine implementing price-time priority and realistic order lifecycle behaviour.
- Supports partial fills, cancellation, cancel-and-replace modification, and trade logging
- Uses ordered bid/ask price levels, FIFO queues, and hash-based active-order lookup
- Includes deterministic throughput benchmarking and documented complexity/design decisions
- Added 10 automated behavioural tests and sanitizer-backed CI
- Processed 100,000 synthetic orders in 22.9 ms average across five identical runs
Tech: C++17 · STL · Data Structures · Algorithms · Make · GitHub Actions
A full-stack learning platform supporting document-grounded assistance, assessments, flashcards, classroom workflows, teacher analytics, and personalised learning roadmaps.
My primary contributions included:
- Implementing a shared AWS Bedrock inference abstraction
- Migrating retrieval, flashcard, assessment, and roadmap workflows to Bedrock
- Adding multimodal document summarisation
- Leading React/FastAPI deployment and service integration on AWS EC2
- Adding 26 offline Bedrock integration tests and full-stack CI
Tech: AWS Bedrock · Python · FastAPI · React · TypeScript · ChromaDB · LangChain · HuggingFace
A team-developed AI honeypot prototype for detecting suspicious messages, engaging potential scammers through a guarded conversational agent, and extracting structured scam intelligence.
My primary contributions included:
- Implementing the guarded conversational agent
- Developing hybrid rule-based and model-assisted scam detection
- Building extraction for UPI IDs, bank accounts, phishing links, and phone numbers
- Adding prompt-injection checks, deterministic fallbacks, response guardrails, normalisation, and deduplication
- Adding 22 offline behavioural tests and Python CI
Tech: Python · GenAI · Information Extraction · Regex · Pytest · GitHub Actions
- Built and validated TypeScript/Supabase backend tools for an AI-powered Teacher Copilot
- Worked with student records, assessment summaries, submissions, and performance analytics
- Developed performance-classification logic using class averages, standard deviation, accuracy, score percentage, and attempt-rate metrics
- Validated retrieval, grounding, preprocessing, recommendation, OMR, and scan-quality workflows
Languages Python · C++ · Java · C · TypeScript · SQL
Machine Learning & Data Pandas · NumPy · scikit-learn · LightGBM · SentenceTransformers · PyTorch · Learning-to-Rank · NLP
Backend & Cloud FastAPI · REST APIs · AWS Bedrock · AWS EC2 · Supabase · SQLite
Core Computer Science Data Structures & Algorithms · Object-Oriented Programming · Database Systems · Concurrency
Engineering Git · Automated Testing · CI/CD · Benchmarking · Reproducible Experimentation
- 100% Merit Scholarship — First Semester, BITS Pilani
- Bronze Honour — International Astronomy and Astrophysics Competition
- Top 10 — K.A.M.A.L.A Mini Hackathon, among 400+ submissions
- Strengthening DSA and core CS fundamentals for technical interviews
- Exploring applied machine learning, search/ranking, and data-driven systems
- Building performance-conscious systems and backend software
- Exploring scientific computing and ML applications on scientific datasets