Machine Learning Data Science

Mahesh Moodukonaje

317-***-**** Indianapolis, USA ***************@*****.*** linkedin.com/in/mahesh-m7/ Portfolio Education

Purdue University Indianapolis, Indiana

Master of Science in Computational Data Science Aug 2023 – May 2025 Ramaiah University of Applied Sciences Bengaluru, India Bachelor of Technology in Electrical and Electronics Engineering Aug 2015 – May 2019 Experience

AI Research Intern, Comcast, Washington DC, United States Jun 2024 – present

• Deployed a containerized Retrieval-Augmented Generation (RAG) pipeline and integrated into Dialogflow with FastAPI to enhance routing capabilities, boosting intent resolution accuracy by 40%

• Fine-tuned a 3B-parameter LLM using QLORA and knowledge distillation from a larger LLM to summarize multi-turn dialogues into single-intent phrases, cutting inference cost by 40%

• Conceptualized a multi-turn intent recognition framework using advanced RAG system (LlamaIndex) for Xfinity Assistant bot, improving intent recognition accuracy by 35%

• Synthesized key findings and authored a research paper to guide the development of robust intent recognition systems in real-world industrial applications (to be published soon)

• Defined 8+ key performance indicators (KPIs) and built a PySpark data pipeline to extract 10,000 high-value records from 10 million records daily, driving operational efficiency Research Assistant, Luddy School of Informatics, Indianapolis, Indiana Jan 2024 – May 2024

• Developed a framework using NLP techniques and machine learning algorithms such as Random Forest to detect depression among dementia caregivers with 70% accuracy

• Analyzed the impact of quality and quantity of noisy data induction in RAG systems, achieving a 15% improvement in information retrieval precision

Data Scientist, Mu Sigma, Bengaluru, India Jun 2022 – Nov 2019

• Led a team of 3 data scientists to develop a customer acquisition framework, performing sentiment analysis and topic identification from 1M+ comments, resulting in customer acquisition across 20+ brands

• Engineered a text classification framework using a Large Language Model (RoBERTa) on Health Department Inspection notes, reducing misclassification by 50%

• Applied LSTM based model to forecast dehydration efficiency and used random forest regression for issue diagnosis, reducing desalter downtime by 4X

• Created ETL pipeline using SQL and PySpark to automate monthly sales reports from 3 data sources

• Built a BERT-based Named Entity Recognition (NER) pipeline to automate lead identification which reduced manual workload of sales team by 30+ hours per month Projects

Job Search Agent FAISS, Langraph, LLM, Docker, REST API Feb 2025 – Mar 2025

• Developed a containerized agentic framework that autonomously scraped job postings from multiple websites and utilized an AI agent to filter and rank listings based on user-specific relevance and preferences Metaphor Detection Python, LSTM, BERT Nov 2023 – Dec 2023

• Engineered an effective metaphor detection framework, employing Bi-directional LSTM with BERT embeddings, achieving an accuracy of 91%

Technical Skills

Languages: Python, SQL, SAS, R, PySpark

Machine Learning Algorithms: Linear Regression, Logistic Regression, Random Forrest, XG Boost, Support Vector Machines, K-means, Time Series Forecasting

Deep Learning: Neural Networks, LSTM, GRU, RNN, ANN, PyTorch, Tensorflow NLP Toolkit: Large Language Models (Llama, Mistral, Qwen, GPT), LangGraph, RAG, llamaindex, Langchain, LORA, QLORA, Knowledge Distillation, Ollama, Spacy, NLTK Cloud: AWS (EC2, S3, Sagemaker), GCP (VertexAI, BigQuery), Azure (Azure Machine Learning, Azure Blob storage, Azure Data Factory)

Tools: Git, DialogFlow, Docker, Databricks, RESTAPI, PowerBI, FASTAPI Libraries: pandas, NumPy, Seaborn, SciKitLearn

Contact this candidate