Skip to Content

Bimal Timilsina

Machine Learning Engineer

I Build ML Systems

About

Innovative Machine Learning Engineer focused on pushing the boundaries of NLP and Large Language Models. Experienced in developing RAG systems, fine-tuning LLMs, and optimizing AI solutions for production environments. Notable achievements include implementing Nepali text-to-speech models and reducing LLM inference latency by over 100% on AWS.

Currently working remotely with US and UK-based companies, building scalable LLM solutions that serve 1000+ concurrent users. Strong foundation in Computer Engineering with expertise in low-resource language technologies. Committed to advancing AI accessibility through practical applications and research.

Looking to collaborate on projects involving multimodal language models, LLM optimization, and innovative NLP solutions. Open to discussing opportunities in AI research and development.

Experience

  1. 2023 - present

    Developed and optimized RAG systems with commercial (GPT-4) and open-source (Llama) LLMs, conducting research, fine-tuning for specific use cases, reducing inference latency on AWS by over 100%, and leading scalable architecture design to support 1000+ concurrent users.

    • Large Language Models
    • RAG
    • Finetuning
  2. Feb - July 2023

    Enhanced email classification accuracy by 15% with DistilBERT fine-tuning and implemented an email automation system for logistics, reducing response time to under one minute.

    • Large Language Models
    • BERT
    • PyTorch
  3. 2022 - 2023

    Developed table structure recognition for bordered, partially-bordered, and borderless tables, enabling cell-level data extraction. Built an invoice parser to extract key details like invoice number, PO ID, vendor, and billing information. Experimented with OCR techniques for document text recognition and conducted research on Nepali language Text-to-Speech.

    • Natural Language Processing
    • Transformers
    • Computer Vision

Projects

  • Background Blurring with Image Segmentation

    This project uses Image Segmentation algorithms to find background and foreground of any image containing person and Blurs the background to give image nice potrait look.

    • Python
    • Tensorflow
    • Flask
    Background Blurring with Image Segmentation
  • Neural Style Transfer

    Implementation of Neural Style Transfer Algoritm to copy style of any image to another content Image which finally outputs a nice styled Image.

    • Python
    • Tensorflow
    • Keras
    • Computer Vision
    Neural Style Transfer
  • My Venue

    My Venue is a venue booking website, where any logged in user can book venues like, party place, pub etc to organize their events. Users can book and confirm the place and pay from website itself, so it provides all in one service.

    • Python
    • Django
    • MySQL
    • HTML
    • CSS
    • JavaScript
    My Venue

Writing