Skip to content
View nhprasan's full-sized avatar

Block or report nhprasan

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.

Maximum 250 characters. Please don't include any personal information such as legal names or email addresses. Markdown supported. This note will be visible to only you.
Report abuse

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

Report abuse
nhprasan/README.md

Namaskaram πŸ™, I'm Prasan

ML Engineer with 1 year of industry experience, 2 years of focused ML training, and 7+ years of cross-functional leadership in non-technical roles. I enjoy building real-world AI systems β€” my work spans LLM-powered conversational AI, computer vision, predictive modeling, and applied robotics.


🧠 Areas of Focus

  • Agentic AI and LLMs β€” LangGraph, LangChain, ReAct agents, tool use, RAG, prompt engineering
  • NLP β€” BERT, SBERT, conversational AI, fine-tuning, parameter-efficient fine-tuning
  • Computer Vision β€” object detection, re-identification, Siamese networks
  • Deep Learning β€” PyTorch, TensorFlow, CNNs
  • Applied AI β€” robotics (ROS 2), healthcare, business intelligence
  • Cloud and Data β€” Azure, GCP, SQL, Pandas

πŸš€ Selected Projects

πŸ“Š BI Agent

Built a conversational BI assistant for a drone services company. It pulls live data from Monday.com via GraphQL, reasons over it using a LangGraph ReAct agent, and lets founders ask plain English questions about pipeline, revenue, collections and sector performance. Deployed on Hugging Face Spaces.

LangGraph LangChain Groq (Llama 3.3 70B) Streamlit GraphQL Hugging Face Spaces

🐯 TigerCensus

A computer vision system for wildlife conservation β€” detects and re-identifies individual tigers from camera-trap images. Extends Microsoft's CameraTraps / Pytorch-Wildlife framework with Siamese networks for one-shot learning. Achieved MAP@3 of 0.8.

YOLO v5 Siamese Networks Transfer Learning Computer Vision

Evaluated a range of ML models for credit risk, with a focus on maximizing recall to catch potential defaulters. Compared Logistic Regression, KNN, SVM, Gradient Boosting and AdaBoost across precision-recall tradeoffs.

Scikit-learn Ensemble Methods Precision-Recall

🩺 MediBot

A context-aware, multi-turn conversational assistant for oncology. Uses a RAG pipeline to ground LLM responses in authoritative medical knowledge rather than hallucinating clinical details.

RAG LangChain Fine-Tuning Vector Database


πŸ§ͺ Background

  • Research Scientist at Humanitarian Technology Labs, working on an ICMR-funded robotic ultrasound system (2025)
  • MS in Machine Learning, University of Arizona (2024)

πŸ“« Find Me Online

LinkedIn Medium Kaggle

Pinned Loading

  1. bi-agent bi-agent Public

    BI agent for analyzing Monday.com Deals and Work Orders

    Python

  2. TigerCensus TigerCensus Public

    This repository extends Microsoft's "CameraTraps" project to tackle individual tiger identification (re-identification or ReID) using deep learning. Building on the "Pytorch-Wildlife" framework, it…

    Jupyter Notebook 1

  3. GermanBankData_LoanDefaultPrediction GermanBankData_LoanDefaultPrediction Public

    The objective was to build a fully reproducible project that uses Machine Learning techniques with the intention of building a prediction model to assess whether the customer of a German bank who h…

    Jupyter Notebook 1

  4. MediBot MediBot Public

    Jupyter Notebook