Research, AI applications, and full-stack systems, each one shipped, not just started.
Research under Prof. Babu Ram Dhawadi: unsupervised detection of Interest Flooding and Cache Pollution attacks in Named Data Networking. One Isolation Forest detector, two independent tracks: miniNDN emulation (PIT-level features) and ndnSIM simulation (face-level tracer counters only). 100% attacker detection with ~0% held-out false positives across tree, dumbbell, and DFN topologies. Paper draft in progress.
Automated roadside speed enforcement built for Nepal traffic police. YOLOv11x detection + ByteTrack tracking, perspective-corrected speed estimation, number-plate reading via PaddleOCR, and automatic evidence capture: vehicle image, plate crop, annotated clip, all into a FastAPI + PostgreSQL backend.
Published Python package: add a semantic-search chatbot to any website
in minutes: pip install ragchatbot. Drop in your docs, run guided
setup, get a grounded Q&A API. Supports Gemini, OpenAI, or local Ollama
backends, no ML experience required.
Character-level encoder-decoder Transformer built from scratch in PyTorch (Vaswani et al.) translating Devanagari Sanskrit to English, with 6+6 layers, 8 attention heads, d_model 512, checkpoint-resumable training with BLEU evaluation, running on Apple Silicon via MPS.
CV pipeline that analyses padel match video in real time: YOLOv8 player detection, TrackNet ball tracking, MediaPipe pose estimation, shot-type classification, and a live annotated scoreboard overlay.
Bilingual (Nepali/English) AI legal assistant giving constitutionally grounded answers for Nepali citizens. RAG pipeline with LangChain + Gemini, FastAPI backend, PostgreSQL document and query store.
Web platform showcasing NLP models for the Nepali language, spoken by 30M+ people yet underrepresented in NLP. React + Flask + Hugging Face Transformers, Dockerised and cloud-deployed.
Civic engagement app connecting citizens with local authorities: notices, grievances, governance transparency. Auth, admin dashboard, English/Nepali support, OCR via Tesseract.js, Docker-containerised services.
Cross-platform app classifying infant cries (hungry, tired, discomfort) from uploaded or recorded audio. LSTM model exported to TFLite, served by a Flask API to a React Native (Expo) frontend with a consistent MFCC pipeline end to end.
Chatbot wired to multiple Model Context Protocol servers: local math tools, a remote expense tracker, and Manim animation generation, all through a LangChain MCP client with a Streamlit UI.