|
A production-style Retrieval-Augmented Generation system with hybrid search (FAISS + BM25), Reciprocal Rank Fusion, and Cross-Encoder reranking. Evaluated on 136 benchmark questions across 1,197 chunks — achieving 98.53% retrieval accuracy with 1.15s average latency.
|
An AI-powered Android health assistant that analyzes symptoms to provide health awareness, preventive tips, and emergency guidance using Gemini AI.
|
|
Android + Windows Desktop productivity system with real-time Firebase sync, secure device pairing, and a full voice pipeline — wake-word detection, speech recognition, and 15+ system commands.
|
Full-stack GitHub analytics platform using rule-based scoring + LLM interpretation to evaluate repo health, risk, consistency, and evolution. Deployed with CI/CD.
|
I build Android apps and AI-powered tools that solve real problems. From a cross-platform voice-controlled productivity ecosystem spanning Android and Windows, to an AI health assistant currently in Google Play Store closed testing — my projects are focused on making technology feel useful, not just functional.
- Final-year B.Tech CSE at Lovely Professional University — Specialization: Machine Learning, Minor: Android Development
- Built OctoVector AI — a RAG system hitting 98.53% retrieval accuracy on 136 benchmark questions
- Currently shipping HAAi to the Play Store
- Open to junior developer roles & internships
- Reach me at midhunkp.work@gmail.com
| Certificate | Issuer | Year |
|---|---|---|
| Supervised Machine Learning | DeepLearning.AI / Coursera | 2025 |
| Android App Development using Kotlin | Meta / Coursera | 2024 |
| Introduction to Large Language Models | Coursera | 2024 |
| Prompt Engineering for ChatGPT | Coursera | 2024 |
| Java Application Development | Lovely Professional University | 2024 |
| Python Programming | Udemy | 2022 |
Open to junior developer roles, internships, and freelance projects.
midhunkp.work@gmail.com