I'm a Computer Engineering student at Thapathali Campus, IOE, Tribhuvan University, focused on Machine Learning, Deep Learning, and Computer Vision — not just to write code, but to solve real problems that matter, especially in a Nepali context.
I like building things from first principles — training models from scratch, understanding what's happening under the hood — rather than just wiring up pretrained APIs. Outside of models and notebooks, I lead teams, coordinate events, and try to build tech that's actually useful to the people around me.
"Code. Learn. Build. Repeat."
Building Nepali Sign Language recognition system — a Transformer classifier built from scratch fused with MediaPipe pose/hand tracking and CLIP-based scene understanding, running as a live inference pipeline.
A six-module automated KYC pipeline — YOLOv8 field detection, OCR matching, face verification, liveness detection, stamp verification, and tampering detection, aggregated into a weighted approval system with a full backend and frontend.
Two-stage computer vision pipeline — YOLOv8 for produce localization, MobileNetV3 for freshness classification — wired into a full-stack app with real-time detection over WebSockets.
Compared nine summarization methods — TF-IDF, TextRank, clustering, supervised classifiers, and BART — on 2,225 BBC News articles using ROUGE metrics and runtime benchmarking.
Transformer encoder-decoder architecture built entirely from scratch in PyTorch — no nn.Transformer module — implementing multi-head attention, positional encoding, and masking, to translate English into Nepali.
Real-time badminton shot classification (forehand, backhand, smash, serve) using YOLOv8 pose estimation and joint-angle trigonometry.
View on GitHubDuplicate-question classifier using Bag-of-Words features with engineered signals; compared Random Forest and XGBoost, reaching 79.45% accuracy.
View on GitHubVolunteer management platform built in a week-long hackathon, with AI-driven volunteer-to-task matching using a Random Forest model.
View on GitHubFull-stack Bagh-Chal (Tiger and Goats) with a Minimax AI opponent, plus local and online 2-player modes with live chat via Django Channels.
View on GitHubClub website with a knowledge-based chatbot powered by PostgreSQL SearchVector for fast query responses.
View on GitHubRC car controlled entirely over WiFi — an ESP8266 in AP mode hosts an HTTP server, driving motors via URL query parameters.
View on GitHubManaged the organization's website and social media presence, and coordinated technical logistics for events.
Leading international partnerships and cross-club initiatives, coordinating joint projects and communication with partner clubs abroad.
Learned WordPress, design thinking, and internet troubleshooting; later facilitated sessions on these topics while collaborating with a global team.
Trained students from Grade 3 to 10 on Canva, Microsoft Excel, HTML, and professional email writing.
When I'm not building models, you'll usually find me...
Open to AI/ML/CV internship opportunities in Kathmandu. Got a project, a question, or just want to collaborate? Reach out.