Bimarsha Khanal

I am a Machine Learning Engineer at Fusemachines in Kathmandu, Nepal, working on document intelligence systems and multimodal LLM-powered agents. My research interests are Multimodal LLMs and AI Agents, and Trustworthy, Privacy-Preserving, and Secure AI Systems.

Currently, I build production AI systems, including document intelligence pipelines for financial workflows handling thousands of documents daily, and research on a multi-modal LLM-powered interview agent (Qwen2-Audio with LoRA fine-tuning and Direct Preference Optimization) to assist hiring managers throughout the recruitment process.

I hold a B.E. in Computer Engineering from Tribhuvan University, Institute of Engineering, Western Region Campus (2019–2024), where I was awarded a full undergraduate scholarship for ranking in the top 3% of the IOE entrance examination. I also completed the year-long Fusemachines AI Fellowship (2023) microdegree in Artificial Intelligence, and I now serve as a Teaching Assistant in the fellowship.

Research Interests

  • Multimodal LLMs and AI Agents (RAG, preference optimization, SFT/DPO, real-time audio processing)
  • Trustworthy AI, Privacy-Preserving, and Secure AI Systems
  • Optimization and resource efficient ML

Current Work

  • Machine Learning Engineer, Fusemachines, Nepal (March 2024 – Present): document intelligence pipelines for financial workflows; multi-modal LLM interview agent research; production ML integration.
  • Teaching Assistant, Fusemachines AI Fellowship (March 2026 – Present): deep learning fundamentals and real-world applications; project mentoring; curriculum development.

See Publications, Research, Projects, Teaching, and my CV for details. You can also download my CV as PDF here.

News & Achievements

  • Aug 2026: Claude Certified Architect - Foundations (Anthropic)
  • 2024: Paddy Disease Detection paper published on arXiv
  • 2024: Delta Hackathon: 1st Runner-Up (Public Littering Detection and Waste Management System)
  • 2023: Leapfrog Revamp Hackathon: 1st Runner-Up (Personalized News Portal); Ethos Hack Cultural Hackathon: Winner; Vertex Hacks Chapter Alpha: Finalist
  • 2022: Techparva Datathon: 1st Runner-Up (Election Outcomes of Nepal)