Available for opportunities
Kadaghari, Kathmandu

Machine Learning & AI

Ankit Katwal

I build end-to-end AI systems — from training language models on Nepali text to production RAG pipelines for governance documents.

GitHub LinkedIn Email Resume
01

About

Ankit Katwal

Hi, I'm Ankit.

I am a Machine Learning practitioner with hands-on experience building end-to-end AI systems — from a GPT-2 style Nepali language model trained on 6.4M rows to a production-ready RAG pipeline for AI governance documents.

Skilled in PyTorch, NLP, and retrieval-augmented systems, I have internship experience in tabular ML and a strong drive for applied AI research. I am currently pursuing a Bachelor's in Information Technology and am always looking for new challenges.

Download Resume Contact Me
02

Projects

Nepali GPT — GPT-2 Style Language Model

Completed Mar 2026
PyTorch
NLP
SentencePiece
Causal Masking
float16
Cosine Decay
  • Built a GPT-2 style Nepali language model in PyTorch with multi-head self-attention, causal masking, a SentencePiece tokenizer, and autoregressive generation
  • Trained on 6.4 million Nepali-text rows and reached ~60 perplexity with stable convergence on an RTX 3050 4GB in ~4 hours
  • Used float16 mixed precision, gradient accumulation, cosine decay, and top-k sampling to fit within GPU memory
View Code

AI Governance & Open-Model Compliance — RAG System

Completed Apr 2026
RAG
Qdrant
PDF Parsing
Dense Retrieval
NLP
  • Built a RAG pipeline for AI governance and compliance documents using layout-aware PDF parsing, hierarchical chunking, and Qdrant vector search
  • Improved retrieval quality with dense-retrieval evaluation, metadata-aware chunking, structured preprocessing, and retrieval precision metrics to refine chunking and embedding choices
View Code
03

Experience

ML Intern

Feb – Mar 2026

Codeveda Technologies · Remote

scikit-learn XGBoost Feature Engineering Git
  • Built tabular classification pipelines using scikit-learn and XGBoost, handling data preprocessing, feature engineering, and model evaluation within a team workflow
  • Iterated on model experiments using Git-based version control, comparing performance across classifiers to select optimal configurations
04

Skills

Languages
Python
C
MySQL
ML & AI
PyTorch
Transformers / NLP
RAG / Qdrant
CNNs / CV
Data & Tools
Pandas
NumPy
Matplotlib
Git
05

Education

Bachelor of Information Technology

Model Institute of Technology, Kathmandu

Sep 2024 – Present

Deep Learning Artificial Intelligence Database Management Computer Networks

+2 Science

Global School of Science

2022 – 2024

GPA 3.57
07

Resume

Your browser does not support embedded PDFs.

Download PDF
06

Leadership & Certifications

College Representative, Code for Change

2026 – Present

Promote Code for Change programs on campus and encourage student participation in social-impact and technology-driven initiatives.

ANAIS
NAAMII
ML Specialization
DeepLearning.AI
CS50x
Harvard University
Intro to Machine Learning
Kaggle
Intermediate Machine Learning
Kaggle
Pandas
Kaggle
Data Visualization
Kaggle
Feature Engineering
Kaggle
07

Contact

Interested in collaborating on ML projects, discussing AI research, or just want to connect? Reach out through any of these channels.

Email
katwalankit615@gmail.com
Phone
+977 9749722813
LinkedIn
ankit-katwal
GitHub
Ankit9424-prog