class MadhuPriya:
def __init__(self):
self.name = "Madhu Priya Pulletikurthi"
self.role = "Data Scientist & GenAI Engineer"
self.education = "M.S. Data Science @ UMBC (GPA: 4.0/4.0)"
self.location = "Ashburn, VA"
self.focus = ["Agentic AI", "RAG Pipelines", "LLM Evaluation", "MLOps"]
self.patent = "India Patent Office - Application No. 202341084582 A"
self.fun_fact = "I turned a dyslexia research project into a published patent 🧾"
def say_hi(self):
print("Thanks for stopping by! Let's build something intelligent together 🤝")
I'm an AI/ML engineer and data scientist passionate about building end-to-end intelligent systems — from raw data pipelines to production-ready GenAI applications. Currently pursuing my M.S. in Data Science at UMBC with a perfect 4.0 GPA, while also serving as a Graduate Teaching Assistant in ML & Data Science.
My work spans Retrieval-Augmented Generation, agentic AI architectures, and LLM evaluation, with hands-on experience from my internship at KPMG India and multiple research-to-production projects.
LangChain · LangGraph · LLaMA 3 · Prompt Engineering · Agentic Tool-Calling
FAISS · ChromaDB · Vector Search · RAG Pipelines · Semantic Search
PyTorch (inference/applied modeling) · XGBoost · NumPy · Pandas · NMF Topic Modeling
Docker · Streamlit · FastAPI · REST APIs
Python · SQL · PostgreSQL · Git · Apache Spark/PySpark · Jupyter
🏫 University of Maryland, Baltimore County — Graduate Teaching Assistant (ML & Data Science) | Oct 2025 – Present
- 🎓 Guided 60+ graduate students through the end-to-end ML lifecycle — feature engineering, model training, hyperparameter tuning, cross-validation, and deployment-readiness across two courses
- 🔍 Reviewed 20+ code and ML submissions per cycle, reducing recurring pipeline errors by ~60% through structured feedback on reproducible ETL patterns and model-validation standards
🏢 KPMG India — AI/ML Analyst Intern | Jan 2024 – Jul 2024 | Bengaluru, India
- ⚙️ Built Python and SQL ETL pipelines integrating Adobe Analytics, Apache Solr, and REST APIs for the Asian Paints client engagement, eliminating ~40% of manual prep and compressing analyst turnaround from days to hours
- 📊 Delivered Adobe Analytics dashboards surfacing customer journey trends and campaign performance for client stakeholders
- 🐳 Standardized ML workflow environments using Docker within a Jenkins-based deployment pipeline across 3 client engagements, reducing re-deployment failures to zero across the engagement lifecycle
| 🏆 Project | 🔧 Tech Stack | 📝 Description |
|---|---|---|
| Voice-Based Agentic AI Interviewer | faster-whisper, LLaMA 3, LangGraph, TTS | End-to-end agentic pipeline (STT → LLM reasoning/tool-calling → TTS) with a dedicated evaluation framework for hallucination detection and failure-mode classification |
| AI Meal Planner (RAG) | RAG, LLaMA-3, FAISS, USDA API | RAG pipeline over 20K+ USDA FoodData records; benchmarked FAISS against 3 alternative vector databases, achieving sub-150ms retrieval on CPU-only hardware |
| Automated News Analytics Pipeline | NMF, TF-IDF, DistilBART, RoBERTa | 12-model NLP benchmark across summarization, sentiment classification, and topic modeling on 300K+ articles; deployed as a sub-30ms Streamlit inference pipeline |
| AI-Interview-Curator | Python, Streamlit, LLMs | Streamlit-based AI tool that analyzes job descriptions & resumes to generate tailored interview questions |
| Dyslexia Diagnosis via ML | ML classifiers, Python | Patent-filed classifier benchmarking 12 ML classifiers, ensemble model achieving 95.2% accuracy — published as "Dyslexia Detection Using Machine Learning," ICACITE 2024 |
| Crime Analysis (Big Data) | Apache Spark, PySpark, Jupyter | Big data pipeline leveraging distributed computing to analyze large-scale crime datasets |
| Optical Flow (RAFT) | PyTorch, RAFT, Computer Vision | Applied a pretrained RAFT model for optical-flow-based motion detection — inference/application-level computer vision |
| YouTube Popularity Prediction | XGBoost, HDBSCAN | ML pipeline predicting pre-publication engagement from YouTube API metadata |
| UMBC Data Science Capstone | Python, ML, Data Science | End-to-end capstone project for UMBC M.S. Data Science program |
| 🏛️ Institution | 🎓 Degree | 📅 Duration | ⭐ GPA |
|---|---|---|---|
| University of Maryland, Baltimore County (UMBC) | M.S. Data Science | Aug 2024 – May 2026 | 4.0 / 4.0 |
| Dayananda Sagar College of Engineering | B.E. Information Science | Aug 2020 – May 2024 | 3.84 / 4.0 |
- 🏆 Google Prompting Essentials — Google (2024)
- 🏆 Google Data Analytics Professional Certificate — Coursera (2024)
- 🤖 Anthropic: Claude 101 (Jun 2026)
- 🤖 Anthropic: Introduction to Claude Cowork (Jun 2026)
- 📜 Published Research: "Dyslexia Detection Using Machine Learning" — ICACITE 2024
- 🔬 Patent Filed: India Patent Office — Application No. 202341084582 A
- 🎓 Perfect 4.0 GPA in M.S. Data Science at UMBC
I'm always open to interesting conversations, collaborations, and new opportunities in AI/ML, GenAI, and Data Science.
"Turning data into decisions, and models into impact."