Skip to content
View Punam918's full-sized avatar

Block or report Punam918

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
Punam918/README.md
Punam Adhikari - AI Engineer

GitHub LinkedIn Portfolio Email

About

I'm an AI and Data Engineer focused on turning complex models and data workflows into reliable, production-ready systems.

My work spans large language models, high-precision retrieval-augmented generation, data engineering, MLOps, DevOps automation, and AI security. I enjoy building systems that are measurable, reproducible, and practical to operate.

Current interests: LLMs · RAG Systems · AI Agents · Data Platforms · MLOps · AI Security

Selected Projects

Project Focus
MCP_Newsletter_Studio MCP-native AI newsletter research and production with source-linked evidence, editorial workflows, and human review
CiteMind-AI-RAGOps Evidence-grounded research assistance with hybrid retrieval, reranking, and transparent citations
Dev-Pilot Human-governed AI debugging with reviewable fixes, controlled tools, approval gates, and automated tests
VerityLake Governed data-to-RAG workflows for ingestion, validation, lakehouse publishing, retrieval, and citation evidence
llm120-from-scratch Decoder-only Transformer training, tokenizer development, evaluation, SFT, and DPO
Adventurework_dataengineering Azure data engineering pipeline covering ingestion, storage, transformation, and visualization
Web_Based_Proctoring Computer vision for head pose, object detection, face verification, gaze tracking, and spoof detection

Research & Engineering

  • Language models: Built a compact decoder-only Transformer from scratch, including tokenizer training, pretraining, evaluation, SFT, and DPO workflows.
  • RAG systems: Design grounded retrieval pipelines with hybrid search, vector databases, reranking, and citation evidence.
  • AI engineering: Build agentic workflows with LangChain, LangGraph, controlled tools, and human approval gates.
  • Data engineering: Work with ingestion, validation, lakehouse publishing, vector indexing, and reproducible data workflows.
  • Production systems: Apply MLOps, CI/CD, containerization, cloud deployment, and AI security practices to make systems reliable and maintainable.

Tech Stack

Python PyTorch TensorFlow Hugging Face LangChain LangGraph FastAPI Apache Spark Airflow Docker Kubernetes Azure AWS PostgreSQL GitHub Actions Git

Building reliable AI systems from data to deployment.

Pinned Loading

  1. CiteMind-AI-RAGOps CiteMind-AI-RAGOps Public

    An evidence-grounded research assistant that combines hybrid retrieval, reranking, and RAG to help users explore scientific papers with transparent citations and verifiable source evidence.

    Python

  2. Adventurework_dataengineering Adventurework_dataengineering Public

    Azure data engineering project including data ingestion, storage, transformation, and visualization.

    Jupyter Notebook

  3. Web_Based_Proctoring Web_Based_Proctoring Public

    A computer-vision-based remote exam monitoring system that analyzes head pose, object detection, face verification, mouth movement, gaze tracking, and face spoofing to help maintain online assessme…

    Jupyter Notebook

  4. Dev-Pilot Dev-Pilot Public

    A human-governed AI debugging workbench that investigates repository failures, proposes reviewable fixes, and verifies changes through controlled tools, approval gates, and automated tests.

    Python

  5. llm120-from-scratch llm120-from-scratch Public

    A compact language-model engineering project that trains a decoder-only Transformer from random initialization on a single laptop GPU, covering tokenizer training, pretraining, evaluation, SFT, DPO…

    Python

  6. VerityLake VerityLake Public

    A governed data-to-RAG platform that connects ingestion, validation, lakehouse publishing, vector indexing, retrieval, and citation evidence through reproducible and reversible data workflows.

    Python