Balu Deokate

About Balu Deokate

With over 13 years of experience in data science and applied research, I am passionate about developing innovative and impactful solutions using cutting-edge technologies and methodologies. My field of expertise is computer and information science, with a focus on AI, machine learning, and computer vision. I currently work as a data scientist at Deloitte USI. At Mobikode, I was responsible for designing, developing, and deploying AI-assistant for chargepoint station service, a project that aims to improve the user experience and satisfaction of EV drivers. I used CNN-based models for station identification and parts detection, as well as various tools and techniques such as PyTorch, YOLO, Python, CV2, LabelImg, and Detectron2. I also collaborate with other researchers and engineers to test, optimize, and scale the AI-assistant system. In addition, I conduct applied research on parts prediction using YOLO, a project that leverages user manuals to extract and analyze data tables and section diagrams. I enjoy working with a diverse and talented team that shares the same vision and mission of creating a more sustainable and accessible future for mobility.

Experience

1. Gen Ai Lead

   Deloitte

   Apr 2024 — Present · Pune, IN

   Project Title: Protocol content generation using Gen AI (LLM and RAG flow) for Life Science ClientProject Details: We generated new protocol using LLM and RAG Flow for Life Science client. We used multiple chains to generate the protocol content with static and dynamic text. Tools and Techniques used in this project are Python, Cluade 3.7 Sonet, AWS Titan as embedding model and OpenSearch as Vector DB, AWS Textract, Azure Document Intelligence Service. Project Title: LLM based chat agent for SQL query optimizationProject Details: We build the chat agent, which takes the user query and generate the SQL query toget appropriate details from SQL database that helps to get different statistics. Tools and Techniques used in this project are Python, GPT 3.5 turbo, Azure Data Lake Storage.Project Title: Advanced workers safety using CNN model and Intel GetiProject Details: Created and used image dataset to train model for workers safety detection usingIntel Geti. Geti platform is used to annotate the data and train the model. I did model optimization bytuning different parameters. Tools and Techniques used used in this are Python, Image Classification, Object Detection, Model Training, Intel GetiProject Title: Pizza quality Inspection using CNN model and Intel GetiProject Details: Created and used image dataset to train model for pizza quality identification for largerestaurant chain. We used multi-label classification model for pizza toppings classification.Tools and Techniques used in this project are Python, Image Classification, Object Detection, Model Training, Intel GetiProject Title: Advertisement Image dataset using GenAIProject Details: Used prompt engineering to develop a high-quality image dataset for advertisement.Tools and Techniques used in this project are Python, LLM, AWS Bedrock, Claude Sonnet, Claude 3 Haiku.

Education

• Savitribai Phule Pune University

  Bachelor of Engineering (B.E.), Computer Engineering

• SJV Daund

  HSC, Science

• University of Ljubljana

  Doctor of Science, Computer and Information Science

• GDKMV Malathan

  SSC, General

• Dr. Babasaheb Ambedkar Technological University

  Master of Technology (M.Tech.), Computer Engineering