MS in CS @ UC Davis | AI/ML Engineer @ CHEST | Software Engineering | LLMs | Explainable AI

I recently finished my Masters in Computer Science at UC Davis. I’m a software engineer building AI applications for hardware. I build LLM-based tools, APIs, and data pipelines that analyze chip designs, detect vulnerabilities, and streamline hardware workflows.

Current Work

• Center for Hardware and Embedded Systems Security and Trust:

  • Set up computing infrastructure to run LLM inference on HPC clusters for hardware code analysis
  • Fine-tuned LLM models to analyze chip design to identify bugs and security vulnerabilities.
  • Co-authored two papers on use of LLMs in Chip Design (ISQED 2025, MLCAD 2025)

Previously:

• Artificial Intelligence Accountability Explainability Lab

  • Designed frameworks that make large language models more accurate and reliable
  • Reduced hallucinations by 30% using RAG and structured reasoning approaches (Chain of Thought Reasoning)
  • Built full stack web application with Python Flask, ReactJS, and MongoDB

• Elite Softwares

  • Built full-stack CRM application with React.js frontend and Flask backend with REST APIs
  • Implemented automated testing with 70% code coverage using PyTest
  • Designed responsive UI that increased user engagement time by 20%
  • Optimized database queries, reducing data retrieval time by 35%

Strengths: backend architecture, API design, performance & reliability, applied LLM systems, hardware-software integration, ML for chip design, ML infrastructure

Interests: AI for adjacent domains, agentic AI, data/infra platforms, full-stack development

• Languages: Python, JavaScript, C++, HTML, CSS
• Databases & APIs: MySQL, PostgreSQL, MongoDB, RESTful API design, OAuth 2.0, JWT, Postman
• Frameworks & Libraries: React.js, Django, Flask, PyTorch, Pandas, Tensorflow, scikit-learn, LLMs
• AI/ML & Vector DBs: Pinecone, FAISS, OpenAI API, RAG, CUDA 12.1, Vector Similarity Search, Transformers
• DevOps & Tools:Github, GitHub Actions, Docker, GCP, JIRA, CI/CD, Selenium

🎮 Gameboi

An open-source generative AI tool that creates complete 2D games from text prompts using GPT-4, DALL·E, and PyGame.

• An open source tool that uses GPT4, Dall-E and PyGame to create customized 2D games based on user prompts.
• It is designed to streamline the game development process by automating various stages, from generating game sprites and assets to writing PyGame code.
• It is also capable of troubleshooting if the game fails to run and improving itself based on user feedback.
• Within a week of open sourcing Gameboi, it had 7 stars and multiple contributions on github.

🧠 StockSense

A local LLM-based financial assistant built by fine-tuning LLaMA3-8B on historic stock market data.

• Fine tuned a Llama3-8B base model on historic stock market data.
• Developed as a part of my Advanced deep Learning course project to create a financial analysis tool which can run locally, analyze and predict stock trends.
• LLMs have the potential to help you understand your financial portfolio based on current news sentiments but you can’t upload your sensitive data onto applications like ChatGPT.
• Observed a 19% improvement in trend prediction and 28% improvement in sentiment analysis over Mistral7B and 10% improvement in trend prediction and 17% improvement in sentiment analysis over base Llama3-8B model

🗣️ HindiSetu

A DeepSeek-powered platform for Hindi learners, currentlybeing used as a experimental learning tool @ Department of Middle East and South Asian Studies, UC Davis.

• Deepseek powered platform to help novice and intermediate learners learn Hindi better.
• Generates transcripts for a given youtube video and generate relevant Q&A based on the transcript which helps build student’s reading and comprehension skills.
• Also implemented a dictionary and word lookup feature to help expand students vocabulary.
• Integrated OpenAI’s APIs and a FAISS vector database to create adaptive assessments, AI generated Q&A pairs, and personalized learning workflows.

1. K. I. Gubbi, M. Halm, S. Kumar, A. Sudarshan, P. D. Kota, M. Tarighat, A. Sasan, and H. Homayoun. Prompting for power: Benchmarking large language models for low-power rtl design generation. In 2025 ACM/IEEE 7th Symposium on Machine Learning for CAD (MLCAD), pages 1–7, 2025.

2. K. I. Gubbi, M. Tarighat, A. Sudarshan, I. Kaur, P. D. Kota, A. Sasan, and H. Homayoun. State of hardware fuzzing: Current methods and the potential of machine learning and large language models. In 2025 26th International Symposium on Quality Electronic Design (ISQED), pages 1–7. IEEE, 2025