Harshit Sandilya

• Led a team of five in developing and deploying an AI-powered Teaching Assistant for a distance learning MBA course, improving student engagement and support. • Designed and implemented CI/CD pipelines for the LLM API infrastructure, automating deployment and scaling of instances using AWS Lambda. • Developed a visualization pipeline to analyze and present question-answer interactions, enhancing the interpretability and effectiveness of the Teaching Assistant.

• Developed 'Neev,' a modular and scalable LLM training framework with optimized data preprocessing, distributed training support, and hyperparameter tuning for efficient model convergence. • Contributed to 'Ved,' a high-performance data pipeline enabling automated extraction, cleaning, and transformation of unstructured web and PDF data for LLM training. • Contributed to the 'Livia' LLM model suite by optimizing model architectures, fine-tuning strategies, and evaluation methodologies to enhance text understanding and generation capabilities.

• Designed and developed robotic solutions for stair climbing and line following using microcontrollers such as Arduino Nano, Raspberry Pi, and ESP32, integrating sensor-based navigation and control algorithms. • Conducted hands-on workshops on robotics, microcontrollers, and machine learning for junior students, covering fundamental concepts and practical implementations. • Mentored a team of ten juniors, guiding them through project development and competition preparation, providing insights in robotics.