Jahnavi Battu

About Me

I'm a Generative AI and Machine Learning Engineer with experience building and deploying AI solutions across Healthcare and Banking domains. My expertise includes Large Language Models (LLMs), Retrieval-Augmented Generation (RAG), Prompt Engineering, NLP, and Machine Learning.

I specialize in developing end-to-end AI systems—from data ingestion and preprocessing to model deployment and production monitoring. I've built scalable RAG architectures using LangChain, LlamaIndex, FAISS, and Pinecone, enabling intelligent document retrieval, semantic search, and knowledge assistants.

Passionate about solving real-world business problems through AI, I enjoy creating scalable, reliable, and production-ready applications using Python, FastAPI, AWS, and modern AI frameworks while collaborating with cross-functional teams to deliver measurable impact.

Education

State University of New York, Buffalo

Master of Science in Engineering Science – Data Science

Jan 2024 – Present

GPA: 3.60 / 4.0

Relevant Coursework: Machine Learning, Big Data Analytics, Statistical Learning, Data Visualization

Indian Institute of Information Technology, Kalyani

Bachelor of Technology in Computer Science and Engineering

Aug 2019 – May 2023

GPA: 3.5 / 4.0

Relevant Coursework: Artificial Intelligence, Database Systems, Computer Networks, Operating Systems

Experience

Machine Learning Engineer – UnitedHealth Group (July 2025 – Present)

Designed and deployed machine learning and Generative AI solutions for healthcare document intelligence, leveraging Python, NLP, Transformer based models, and Retrieval-Augmented Generation (RAG) pipelines to automate information extraction from clinical and administrative documents.
Developed end-to-end ML pipelines for processing unstructured healthcare data, including document ingestion, text preprocessing, embedding generation, vector search, and LLM-based response generation using frameworks such as LangChain, Hugging Face, and FAISS and Pinecone.
Fine-tuned and evaluated NLP models for healthcare use cases including clinical entity extraction, document classification, medical text summarization, and question-answering, improving accuracy and reducing manual document review efforts.
Built and containerized scalable ML inference services using FastAPI and Docker, deploying AI workflows on AWS infrastructure including S3, EC2, and SageMaker to integrate machine learning models into healthcare applications.
Implemented model evaluation and monitoring workflows using metrics such as precision, recall, F1-score, hallucination evaluation, and retrieval accuracy to ensure reliable AI outputs in healthcare applications.
Collaborated with data engineers, healthcare analysts, and product teams to optimize AI workflows, improve system performance, and deliver production-ready ML solutions supporting healthcare operations.

Data Scientist – Amgen (June 2021 – Dec 2023)

Designed and implemented machine learning–based demand forecasting models using Python, SQL, XGBoost, and Prophet to predict weekly pharmaceutical product demand across multiple regions, improving forecast accuracy by 18%, supporting proactive supply planning.
Built scalable data pipelines by integrating sales, inventory, pricing, promotional, and market datasets, engineering time-series features including lag variables, rolling statistics, and seasonal indicators to enhance model performance and forecast reliability.
Conducted exploratory data analysis and feature importance evaluation to identify key demand drivers, enabling commercial teams to make informed decisions on product launches, promotional strategies, and inventory allocation.
Collaborated with commercial analytics, supply chain, finance, and business stakeholders to translate forecasting insights into actionable recommendations, reducing stockout risk by 15% while optimizing inventory levels across multiple product portfolios.
Automated model training, validation, and forecasting workflows using Python, SQL, and AWS, ensuring reproducible pipelines and consistent delivery of demand forecasts for business planning and operational decision-making.
Developed interactive Power BI dashboards and executive reports to monitor forecast accuracy, regional demand trends, and key business KPIs, enabling leadership to evaluate forecast performance and support strategic planning initiatives.

Skills

Generative AI & LLMs

GPT, LLaMA, RAG, Prompt Engineering, LLM Evaluation

LLM Frameworks

LangChain, LlamaIndex, HuggingFace Transformers

Vector Databases & Search

FAISS, Pinecone, Semantic Search, Hybrid Search

Machine Learning

Classification, Regression, Clustering, Feature Engineering, XGBoost, Random Forest, Logistic Regression

Natural Language Processing

NER, Text Classification, Sentiment Analysis, Document Summarization, Information Extraction

Programming Languages

Python, SQL

Data Processing & Analysis

Pandas, NumPy, Data Cleaning, EDA, Feature Engineering

Cloud & Deployment

AWS (S3, Lambda, API Gateway), Docker, FastAPI, REST APIs, CI/CD

Databases

MySQL, PostgreSQL

Visualization & BI

Power BI, Tableau, Matplotlib

Tools

Git, GitHub, Linux, VS Code, Jupyter Notebook

Projects

ICU Vitals Forecasting

LSTM model for ICU vitals forecasting with patient-level split.

PneumoVision

CNN-based pneumonia detection with Grad-CAM and Streamlit inference..

GenNarrate

LLM app for script generation and dialogue refinement.

Resume Reviewer using LLMs

Offline resume-review system using Zephyr-7B with Llama.cpp, PyMuPDF, and Streamlit to extract PDF text, compute ATS scores, and generate real-time, high-quality feedback.

Deep Audio Classifier

TensorFlow-based classifier with 92% accuracy.

CyberBullying Detection

NLP-based detection model with a focus on identifying cyberbullying content.

Plant Disease Detection

Built CNN model achieving 95% accuracy in identifying plant diseases.

Smart Voting System

Blockchain-based voting system with enhanced security features.

Titanic Survival Prediction

Automated ML pipeline, Dockerized for real-time predictions.

EmojiMapper

Text-to-emoji mapping using NLP techniques for a fun interactive experience.

Research & Publications

Spatiotemporal Forecasting of Meteorological Variables Using ConvLSTM Networks

Research Area: Deep Learning • Time Series Forecasting • Climate Analytics • Computer Vision

This research investigates the application of Convolutional Long Short-Term Memory (ConvLSTM) networks for forecasting meteorological variables such as temperature, humidity, and atmospheric conditions using spatiotemporal climate datasets. The proposed approach leverages both spatial and temporal dependencies to improve forecasting accuracy compared to traditional machine learning and statistical methods.

Key Contributions:

Designed and implemented a ConvLSTM-based deep learning architecture for multi-step weather forecasting.
Processed and analyzed large-scale meteorological datasets to capture spatial and temporal weather patterns.
Improved forecasting performance by integrating spatial feature extraction with temporal sequence modeling.
Evaluated model performance using multiple forecasting metrics and compared results against baseline approaches.
Demonstrated the effectiveness of deep spatiotemporal learning techniques for climate analytics and environmental forecasting applications.

Jahnavi Pravaleeka Battu