AI/ML Engineer
I build intelligent systems — from data-driven models to scalable ML pipelines. One years delivering AI solutions that learn, adapt, and perform at scale.
# AI/ML Engineer: Pukar Rimal from brain import creativity, logic, data import numpy as np import tensorflow as tf stack = ["Python", "ML", "DL", "NLP", "LLMs"] def ship(idea): return f"{idea} → deployed!" print(ship("AI systems that adapt"))
I'm an AI/ML engineer based in Kathmandu, passionate about building intelligent systems that learn, adapt, and solve real-world problems. I specialize in crafting models, pipelines, and solutions that scale seamlessly.
My philosophy: AI is future, lets be part of it. I develop models that are interpretable, pipelines that are robust, and systems that deliver impact efficiently. I integrate the build model in the system reducing hallucinations and delivering desired output
When I'm not training models or exploring datasets, I'm experimenting with new AI frameworks, reading about neural networks, or exploring ways to blend AI with creative projects.
Fine-tuned Unsloth LLaMA-3.1 on domain-specific datasets to enhance reasoning and context understanding. Achieved improve response accuracy, faster inference, and robust performance on specialized NLP tasks.
Developed a Retrieval-Augmented Generation (RAG) legal assistant app, leveraging Qdrant as the vector database for fast, context-aware document search. Enables users to query legal texts and receive accurate, AI-generated guidance instantly. It is based on Nepal Constitution.
Built a Retrieval-Augmented Generation (RAG) app that ingests PDFs as input and provides accurate, context-aware answers. Leveraged ChromaDB for efficient vector storage and fast retrieval, enabling users to query documents seamlessly.
Developed an intelligent job portal that analyzes user CVs and recommends relevant jobs using AI algorithms. Features real-time search, skill-based matching, and personalized career guidance powered by ML models.
Built a CNN-based image classifier that distinguishes between nature and city images. Trained on diverse datasets, achieving high accuracy and robust performance for real-world image classification tasks.
Built a system that automatically analyzes news content to detect misinformation and flag potentially fake articles. Designed for accurate, fast, and reliable classification of news items.
Built multi-level sales forecasting models using Decision Trees and ensemble models including Random Forest, XGBoost, and LightGBM, applying feature engineering techniques to improve model accuracy. Developed an ERP chatbot using Retrieval Augmented Generation (RAG) to provide context-aware responses from enterprise data. Wrote and optimized multiple SQL queries for handling ERP databases, report generation, and data analysis. Created REST APIs using FastAPI and Django, developed backend services, and integrated the APIs into an ERP website using Next.js.
Built Retrieval Augmented Generation (RAG) systems integrating Large Language Models (LLMs) with external knowledge sources for accurate, context-aware outputs, and developed Artificial Intelligence (AI) driven recommendation systems for personalized content. Built and deployed Optical Character Recognition (OCR) applications for automated text extraction, and designed web crawling pipelines with automated pagination, workflow automation, and vector embeddings for semantic search. Applied prompt engineering to improve Large Language Model (LLM) response quality and reduce hallucinations, developed Representational State Transfer (RESTful) APIs using FastAPI, and containerized and deployed AI systems using Docker on cloud infrastructure.
Researched and evaluated web crawling and scraping strategies for dynamic, large-scale websites, and designed scalable Crawl4AI pipelines for high-volume data extraction with pagination, session handling, and workflow automation. Cleaned and normalized unstructured web data using LLM-based workflows and NLP techniques, performed unit and end-to-end testing using pytest, and applied prompt engineering to improve Large Language Model (LLM) response quality and reduce hallucinations.
Open to AI/ML & software engineering roles and interesting side projects.