AI/ML Engineer and Technical Project Manager with 4+ years building production systems in Machine Learning, LLMs, Recommender Systems, and cloud-native AI infrastructure. Former AI Team Lead at Inflancer Technology; now building independently across AI, tooling, and open-source. Ranked #61 among Nepal's top open-source contributors on GitHub. Actively seeking ML Engineer or Technical PM roles.
- Architected and deployed production Recommender Systems on Docker and Kubernetes, directly improving content personalisation and user engagement at scale.
- Built NLP embedding-based semantic search to replace keyword lookups with intent-aware, contextually relevant content discovery.
- Designed scalable AI microservices integrating Kafka for real-time data pipelines and event-driven processing on cloud infrastructure.
- Led cross-functional AI collaboration across product, engineering, and business — translating model capabilities into measurable outcomes.
- Drove end-to-end delivery of AI and product initiatives using Agile methodologies, owning sprint planning, risk mitigation, and stakeholder alignment.
- Coordinated cross-functional teams to maintain on-time, high-quality releases with clear communication up to leadership.
- Fine-tuned a Llama-2 7B model on a custom code-description dataset, shipping a fully functional Python code generator.
- Built a real-time hand gesture recognition system with CNNs and OpenCV, deployed live via Vercel.
- Contributed to ML research on Nepali Music Genre Classification and Multi-modal Hate Speech Detection, supporting a mentor's M.Sc. thesis.
- Built credit card fraud detection and customer churn prediction models; achieved 86.55% accuracy with Random Forest, SMOTE, and hyperparameter tuning.
- Delivered an SVM spam classifier at 98% accuracy and an NLP movie genre classifier with a production-ready interface.
- Mentored students through ML fundamentals, live debugging, and model improvement during a week-long intensive training camp.
Zero-dependency, physics-powered AI knowledge graph engine with interactive bubble-graph visualization. Explores AI concept relationships dynamically — no external libraries required.
Zero-backend quiz engine — pure HTML/CSS/JS/JSON. 50 randomised questions, keyboard nav, dark/light mode, session resume, rich results breakdown. Fully open-source for educators.
Natural language-to-Python code generator powered by a fine-tuned Llama-2 7B model. Built with ReactJS and deployed on Vercel.
NLP sentiment classification using DistilBERT, achieving 89% accuracy on benchmark datasets.
CNN + OpenCV pipeline for real-time gesture-to-sign-language translation. Built for accessibility applications.
AI-augmented variant with adaptive difficulty and enhanced game mechanics — built to explore algorithmic game logic and agent-driven behaviour.
- #61 in Nepal – Top GitHub Committers: Ranked among Nepal's most active open-source contributors on the global top.committers leaderboard.
- Tech Writer on Medium: Weekly articles on LLMs, AI engineering, and software tools — reaching 1K+ readers/week. Profile ↗
- Industry Case Studies: Independently researching how AI-driven companies (Spotify, Netflix, Airbnb) architect and monetize their ML systems — insights applied to engineering and product decisions.
- Open-Source Education: Maintains The Coder's Handbook, how-I-learned-python, and how-I-learned-ml as structured beginner guides.
- Deltathon Organizer: Managed logistics and coordination for the ACES major hackathon; later competed as Team ZAV.