What I bring

AI that works beyond the demo.

Practical experience across the full path from unstructured input to dependable production workflows.

01 / AI systems

From documents to decisions

Multimodal pipelines that extract financial intelligence from broker emails, PDFs, and images.

02 / Reliability

Useful, controlled AI tools

Authenticated MCP access, typed tools, structured outputs, and deterministic validation.

03 / Data engineering

The infrastructure underneath

Real-time Python and AWS pipelines that deliver financial data for downstream analytics.

Explore my experience

Selected projects

Applied ideas. Concrete applications.

Applied LLMs

Customer Ticket Classification

Used Mistral-7B-Instruct for translation and a hosted LLM for ticket categorization, exposing API endpoints for CRM integration.

Multilingual support workflows

  • Mistral-7B-Instruct
  • LLMs
  • API integration
Document intelligence

Retrieval-Augmented Generation (RAG) Chat Application

Built a Streamlit PDF question-answering application using FAISS for semantic retrieval and document-grounded responses.

Answers grounded in PDF content

  • Streamlit
  • FAISS
  • RAG
Natural language processing

Nepali Plagiarism Detection

Developed a Nepali POS tagger and plagiarism-detection system using a Nepali Samdakosh synonym set, implemented for TUCL’s Nepali thesis collection.

Nepali-language text analysis

  • NLP
  • POS tagging
  • Synonym matching
Explore the projects