$ cat welcome.txt
Utsab Dahal उत्सब दाहाल
AI/ML Engineer · Backend Systems · Open Source
Pulchowk Engineering Campus, Nepal
I design AI systems that are useful in production, from retrieval and evaluation to the API and deployment layer.
Electronics engineer from IOE Pulchowk Campus, Nepal. My work spans retrieval-augmented generation, agents, world models, and dependable backend services.
Fusemachine AI Fellow 2025 and Samsung Innovation Campus participant. I contribute fixes, tests, and documentation to open-source projects including Microsoft RD Agent, scikit-learn, Keras, pandas, and PR-Agent.
{ "languages": ["Python", "C++", "JavaScript", "C"], "ai_ml": [ "Transformers", "LangChain", "LangGraph", "RAG", "AI Agents", "Scikit-Learn", "TensorFlow", "PyTorch", "Pandas", "NumPy", "Matplotlib" ], "deep_learning": [ "Neural Networks", "RNN/LSTM", "NLP", "Computer Vision", "OpenCV" ], "backend": [ "FastAPI", "Docker", "AWS", "Git", "Linux" ] }
Define the user, data flow, failure modes, and operating constraints before choosing a model.
Pair model quality with tests, evaluation sets, and useful signals from the real workflow.
Turn experiments into documented APIs, repeatable deployments, and maintainable code.
JEPA-style latent dynamics model that predicts constrained clinical state trajectories.
A language model trained from scratch on a Nepali text corpus.
An agent workflow for detecting manipulated media and coordinating analysis steps.
Document-grounded question answering with retrieval and a focused user workflow.
$ ./send_message.sh
Type help for available commands.