I am a Computer Engineering graduate with a Minor in Data Science from IIT Mandi (CGPA: 8.39). I specialize in bridging the gap between artificial intelligence systems and interactive virtual experiences. My focus is on Machine Learning / Deep Learning (Explainable AI, RAG frameworks, Neural Decoding) and Game Development (Godot Engine, Unity, gameplay programming, physics-based simulations, and custom shaders).
- Machine Learning: Researching EEG visual digit decoding, privacy-first healthcare retrieval systems (RAG), and Explainable AI (XAI).
- Game Development: Architecting gameplay systems, AI behaviors, physics-based simulations, and graphics shaders in Godot Engine and Unity.
- Published Researcher: Author of an EEG classification model published in DIECAI-2025.
Machine Learning & Data Science
 
-teal?style=flat-square){kind=icon}
|
Game Development & Languages
 
|
- MedRAG: Privacy-First Clinical Document Intelligence | Python, BioBERT, ChromaDB, Llama 3
- Designed a localized, offline-first clinical RAG framework achieving an 82.3% retrieval hit rate with sub-minute query latencies.
- Secures sensitive patient records by operating fully within local network infrastructures.
- Live Demo
- Adaptive Dual-Tier Intrusion Detection System for IoT | Python, XGBoost, PyTorch, SHAP, LIME
- Created an edge-cloud collaborative network IDS utilizing decision trees at the edge and Deep ResNet at the cloud.
- Achieved 99.98% classification accuracy and audited decision features using SHAP/LIME.
- Live Demo
- Data-Centric Neural Decoding of Visual Stimuli from EEG | Python, PyTorch, Scikit-learn
- Developed a representation learning pipeline for 10-class visual digit decoding from noisy 14-channel EEG recordings.
- Filtered trials using Density Peak Clustering (DPC) and utilized Triplet Margin Loss to achieve 100% classification accuracy.
- Rice Quality AI & Crop Yield Prediction | Python, YOLOv8, ResNet, FastAPI, Docker
- Built a YOLOv8 defect classifier (99.3% accuracy) and ResNet-18 neural regression framework.
- Implemented Monte Carlo Dropout for uncertainty estimation and deployed using serverless FastAPI on Hugging Face.
- Live Demo
- Decagon | Godot, GDScript, Puzzle Design
- An elegant, shape-merging puzzle game featuring clean mechanics, scoring thresholds, and minimalist vector designs.
- Play on itch.io
- Spin OverDrive | Godot, GDScript, Physics Programming
- A browser-based arcade fighting game simulating fast-paced spinning top (Beyblade) arena combat.
- Engineered custom physics equations to model rotational drag, inertia, and elastic collisions between tops.
- Play on itch.io
- Accurate Classification of EEG using a Lightweight Two-Stage Model with PaCMAP + K-NN and XGBoost
- Published in Advances in Electronics, Communication and Artificial Intelligence (DIECAI-2025)
- ISBN: 978-81-7770-253-8
- Researched high-efficiency EEG classification combining advanced manifold learning (PaCMAP) with ensemble models.