Dhruv Belawat

Software Intern @Texas Instruments

Bangalore, IN

Worked on automating MCAL (Microcontroller Abstraction Layer) generation for new devices using large language models (LLMs) such as Claude and Copilot, without fine-tuning or modifying the models.Key contributions included:Developed a full-stack, one-click solution that automates the entire workflow, eliminating the need for manual prompt entry and making the process seamless.Implemented techniques like prompt engineering, chain-of-thought prompting, retrieval-augmented generation (RAG), and feedback loops to improve accuracy, reduce hallucinations, and manage large input sizes.Designed a solution pipeline where device specifications and existing MCAL code are processed by LLMs to generate new MCAL code, followed by automated quality checks and iterative feedback.Conducted extensive experiments to benchmark accuracy, response quality, and effort trade-offs across prompting and augmentation techniques.Impact: Demonstrated the potential of LLMs to significantly reduce manual effort in embedded software development by accelerating MCAL generation for new devices.