name: Sumit Shrestha
located_in: Earth 🌏
current_focus: AI, Machine Learning & Intelligent Web Apps
research_interests: [LLM agents, RAG systems, applied NLP, explainable AI]
learning: LLMs, Neural Networks, Model Deployment & MLOps
motto: "I learn by doing, breaking, and building again."
fun_fact: "I debug better with lo-fi music and cold tea ☕"I began with web development, and lately I've been going deeper into AI: machine learning, LLM-powered tools, automation, and practical systems that make software feel smarter. My current goal is to connect clean web experiences with intelligent features that are actually useful — with a particular interest in multi-agent systems, retrieval-augmented generation, and making AI outputs verifiable and explainable. 📫 Reach me: LinkedIn
|
A full-stack autonomous research platform where specialized LLM agents (planner, web researcher, document researcher) decompose complex questions, retrieve evidence via Hybrid RAG, and validate claims with adversarial fact-checking — producing source-backed, verifiable reports.
|
Capstone project: a retrieval-augmented medical Q&A chatbot grounded in curated medical literature, designed to reduce hallucinations and surface supporting sources for every answer.
|
|
A multilingual, multimodal (text / voice / image) mental-health support assistant using hybrid retrieval over curated psychoeducational resources — empathetic responses with source attribution, built with safety disclaimers first.
|
A hybrid AI system combining a skill ontology + rule-based reasoning with semantic similarity (SentenceTransformers) to analyze resumes against job descriptions and produce explainable, human-readable feedback.
|
|
An NLP document assistant that generates extractive summaries and extracts key terms from PDF/DOCX files through an interactive web UI.
|
End-to-end MLOps workflow demonstrating reproducible training, testing, and deployment practices — structured project layout, experiment tracking, and CI-friendly tooling.
|
|
|
|
|
