Kiran Bodipati

GPA: 4.0/4.0

Some relevant coursework includes Regression, Advanced Computer Vision, Bayesian Modelling, Advanced Artificial Intelligence, Advanced Database Systems, Science of LLMs, Graph Theory etc.

Some relevant coursework are Data Science, Machine Learning, Data Structures and Algorithms, Linear Algebra, Calculus, Probability Theories, Statistics, OODP, Databases, Artificial Intelligence, Machine Learning, Software Engineering, Deep Learning & Neural Networks, Data Mining, NLP, Regression Analysis, Computational Finance, Data Visualisation, Product Science, Time series analysis, Information Retrieval.

I also have a Minor in Business.

Here, I build technology driven real estate solutions. A prominent project is the deployment of an AI based land analytics system. I've also built a Face Identification model using PyTorch to count users walking in and out of a location to provide analytics of customer behaviours in in the location, which can improve customer hospitality, in locations that have stores like retail etc.

I'm also involved in business development, business expansion strategies and customer acquisition strategies.

Developed dynamic scraping algorithms using Python, Tesseract, and IndicTranslate to extract over 3 million historical revenue court cases, digital land records and property transactions from government court records and SOLR to index them. Developed and took independent responsibility for a Graph network based document validation system to validate the completeness of a title chain and predict flag potential missing links in the chain with an accuracy of 91%.

Performed Root cause analysis for fault detection in manufacturing processes using Bayesian Belief Networks to reduce downtime by 15%. Developed pre-processing tool of automation pipeline, incremental learning algorithms to update the BBN and leveraged Pytest to unit test the tool, and FastAPI, Streamlit & Plotly to deploy the user-interface.

Developed a general brand detection model pipeline to identify logos in daily objects present in product listings and classify brands by employing the OpenCV, YOLOv5 and Pytorch Frameworks. Employed image pre-processing techniques, produced custom training edge cases, finetuned open-source CNN models, etc to improve recall & precision from 0.1 to 0.35 & 0.6 respectively (150 class model)

Designed end-to-end video game credits extraction service using OCR model to extract names and job titles of various people working on the game. Created a Flask and React based website to bind profiles of people listed in-game to Linkedin profiles, to deliver a real time search repository of game developers.