About Me

I am a Computer Engineering graduate from Institute of Engineering, Pulchowk Campus. Currently, I am a research intern at NAAMII and Freelance AI/ML developer at Upwork. With a solid foundation in Natural Language Processing, Large Language Models, and GenAI, I focus on developing innovative solutions while staying updated with the latest advancements in AI.

My research interests focus on two critical challenges: enhancing the transparency and reliability of LLMs, and ensuring they are truly inclusive, serving diverse linguistic and cultural communities. I'm actively looking for Research opportunities.

email: [email protected]

Degree: Bachelors in Computer Engineering

address: Kathmandu, Nepal

Experience

NLP Research Intern

NepAl Applied Mathematics and Informatics Institute for Research (NAAMII)

August 2024 - Present

  • Conducted research on Hate Speech Detection in Devanagari Scripts including Nepali and Hindi.
  • Finetuned different multilingual models, including XLM-RoBERTa, MuRIL, and IndicBERT, with labeled datasets from Shared Task on CHiPSAL, COLING.
  • Employed advanced data augmentation techniques such as contextual filtering with BERT cosine similarity and LLM-based augmentation to handle class imbalance.

Freelance ML/AI Developer

Upwork | Remote

July 2024 - Present

  • Authored tutorials on cutting-edge technologies in NLP, including Retrieval-Augmented Generation (RAG) with LLaMA-3.1, Graph RAG, Binary Quantization, LLaMA-3 and Phi-3 architecture, PyTorch, and LLamaCPP.
  • Course Developer for Data Science Course aimed at career-switchers, covering Python, Pandas, NumPy, and Statistics.
  • Implemented a system for auto-evaluation of descriptive answers based on defined rubrics (Grammar, Vocabulary, Content, etc.) using Agentic RAG approach with Llama-3.1 and LangChain.

Junior Machine Learning Engineer

ICEBRKR | Lalitpur, Nepal

April - July 2024

  • Developed NLP systems to enhance the app's messaging features, including fine-tuning T5 model to generate dynamic chat phrases and comparing performance with different LLMs.
  • Managed the complete machine learning pipeline from dataset creation to FASTAPI endpoint deployment. Collaborated with the backend team to integrate these endpoints into the overall system architecture.

Backend Engineer Trainee

UXCam | Kathmandu, Nepal

May 2023 - July 2023

  • Developed RESTful APIs using Django REST Framework, conducted API testing with Postman, and utilized PostgreSQL and MongoDB for database management.

Publications

2024

NLPineers@ NLU of Devanagari Script Languages 2025: Hate Speech Detection using Ensembling of BERT-based models

Anmol Guragain*, Nadika Poudel* (* equal contribution) , Rajesh Piryani, Bishesh Khanal

Accepted at COLING 2025

Skills

Python
PyTorch
Pandas
Numpy
Matplotlib
LangChain
FastAPI
Django
SQL
MongoDB
AWS
GCP

Education

Bachelors in Computer Engineering

Tribhuvan University, Pulchowk Campus

2019-2024 | Aggregate : 82.80 % (A+)

- Relevant Courses: Human Language Technology, Bigdata, Enterprise Computing,DSA, COA,AI, Operating System, Distributed System, Computer Network, Computer Graphics, Microprocessor, OOP, C programming

My Projects

Grievance Recognition using Nepali Text/Speech

Grievance Recognition using Nepali Text/Speech

Classifies Nepali voice and text complaints received on Nepal Government portals. Finetuned Wav2Vec2 for ASR and Nepali-BERT model or multi-class text classiication.

Medical Jargons Simplification App

Medical Jargons Simplification App

Translates Medical reports and texts into simple terms using fine-tuned T5 model and LLM(LLama3). Implemented chatbot using RAG to explain medical jargons with LangChain and Llama3.

Question and MCQ Generation

Question and MCQ Generation

Generated Questions and MCQs from text paragraphs by fine-tuning T5 model.

AI Yoga Trainer

AI Yoga Trainer

AI yoga trainer that provides real-time audio feedback to correct yoga pose.

Drishya

Drishya

A smart goggles that detects its surroundings through a camera-fed AI model that informs users with very low or no eyesight by prioritizing the objects and informing their positions.

Newsly

Newsly

A mobile app that provides summarized and audio news, news according to categories, and news on the platform you love.

My Blogs

Advanced RAG

Advanced RAG Implementation using Hybrid Search, Reranking with Zephyr Alpha LLM

Let’s say with the limitations of LLMs:...

Read Blog
Finetuning Zephyr LLM

Finetuning Zephyr 7B with QLoRa and PERF for Customer Support Chatbot

Here, I am talking about how I trained a quantized Zephyr-7B-beta LLM model using Google Colab’s free tier T4 GPU ....

Read Blog

Achievements

Achievement 1

Generation Google Scholarship 2022

Awarded by Google as one of 55 scholars in the APAC Region for demostrating strong passion in technology,exemplifying leadership and strong academic performance.

Achievement 2

World Finalist at Microsoft Imagine Cup 2023

Selected to compete at the World Finals of Imagine Cup 2023 by Microsoft with project Drishya

Achievement 3

National Representative- Seeds for the Future 2022

Selected as one of 7 finalists to represent Nepal at Thailand for 11 days bootcamp in AI,5G technology, Cloud Computing and E-governance.

Achievement 3

Hult Prize at IOE, Pulchowk Campus

On-campus Winner

Achievement 3

Hitachi Technergy - Winner

Won 5 days hackathon on energy forecasting category organized by Hitachi Energy.

Achievement 3

Naamche Hack-A-Week - Winner

Won 7 days hackathon organized by Naamche under Health Category.

Achievement 3

Leapfrog Revampathon - Winner

15 days hackathon on revamping Nepali news portal.

Achievement 3

Ace-Ignite Hackathon - Runner Up

48 hours hackathon by Ace College

Achievement 3

UTEC Hackathon - Winner

36 hours hackathon by UTEC College

Achievement 3

Codecamp by UXCam - Winner

2 days hackathon by UXCam and internship placement