Nitesh Nepal

1.Kidney Disease Detection Using MLflow and DVC
GitHub Repository: https://github.com/NiteshNepal/Kidney_Disease_Detection_Using_MLflow_DVC

Developed an end-to-end machine learning pipeline for kidney disease detection, emphasizing reproducibility, scalability, and efficient experiment tracking. Key aspects of the project include:

• Data Version Control (DVC): Implemented DVC to manage datasets and model versions, ensuring consistent and reproducible experiments.
• MLflow Integration: Utilized MLflow for tracking experiments, logging metrics, and managing model artifacts, facilitating seamless experiment management.
• Modular Pipeline Architecture: Designed a modular codebase with clear separation of concerns, enhancing maintainability and scalability.
• Configuration Management: Employed YAML configuration files to manage parameters and settings, allowing for easy adjustments and experimentation.
• Automation: Automated the workflow from data ingestion to model evaluation using DVC pipelines, streamlining the development process.

2.Agentic Workflow for Autonomous Coding AI Agent
GitHub Repository: https://github.com/NiteshNepal/coding_AI_agent_using_PydanticAI_and_Langraph

Developed a Coding AI agent that focusing on creating an autonomous system capable of generating Pydantic AI agents. Key contributions include:

• Integration of LangGraph: Implemented LangGraph to establish a comprehensive agentic workflow, enhancing the system's ability to manage complex agent creation processes.
  
• Reasoning LLM Utilization: Employed reasoning language models (e.g., O3-mini or R1) to analyze user requirements and documentation, facilitating the creation of detailed scopes for agent development.
  
• Specialized Agent Coordination: Designed specialized coding and routing agents guided by the generated scopes to produce high-quality Pydantic AI agents.
  
• Documentation Crawling and RAG System: Developed an intelligent documentation crawler and Retrieval-Augmented Generation (RAG) system using Pydantic AI, LangGraph, and Supabase to build other Pydantic AI agents.
  
• Local LLM Support: Integrated support for local language models with Ollama, enhancing the system's flexibility and accessibility.

3.LLMs-from-scratch – Building a GPT-like Language Model
GitHub Repository: https://github.com/NiteshNepal/building-Your-Own-Custom-LLM

Contributed to the development of a GPT-style large language model (LLM) using PyTorch. Key aspects of the project include:

• Transformer Architecture Implementation: Developed core components such as multi-head self-attention, positional encoding, and layer normalization to construct a functional transformer model.
  
• Model Training and Fine-Tuning: Executed the training pipeline, including data preprocessing, tokenization, and optimization, to pretrain and fine-tune the LLM on sample datasets.
  
• Educational Codebase Contribution: Enhanced the repository's educational value by contributing clear, well-documented code and Jupyter notebooks, facilitating learning and experimentation for others.
  
• Experimentation and Evaluation: Conducted experiments to assess model performance, adjusting hyperparameters and architecture components to optimize results.

• IBM Data Science
• Microsoft Data Science
• Udemy Machine Learning
• Google Cloud Study Jam

• Actively participated in Rotary Club
• Interact Club
• Red Cross Society
• Multiple District Level Sports Medals
• Fashion Club