Shivam Naik

- Conducted safety severity analysis via Sentiment Analysis using LLMs on descriptive data of road activity. - Improved the performance by 10% by fine‑tuning the BERT model to suit the road activity description. - Design a segmentation pipeline for ADAS vehicles for German roads using TensorFlow and Python. - Finetune the open-sourced SegmentAnything model to robustly segment road surfaces automatically, improving segmentation efficiency by 18%.

• Streamlined an angular-based admin platform for an object-detection based garden watering system, yielding a 40% reduction in water wastage and a 20% increase in overall user satisfaction. • Collaborated with external stakeholders to define coding standards and project timelines, improving customer satisfaction by 16%. • Conducted data analysis of soil sensor data, uncovered prediction strategies, and improved automation by 8%.

• Fabricated a Python-based depth-enabled hand detection algorithm; integrated and showcased in CES 2019 as The PAI-Projector. • Translated the code to C++ for embedded integration, reducing the execution time from 18 to 3 seconds. • Attained 85% of precision at the real-time simulation of a finger touch on calibrated surfaces, enhancing the usability of kitchen-based assist appliances.