VIGNESHWARAR JEGATHEESAN

– Designed an early evaluation pipeline using LLM-as-judge methodology to measure factual accuracy and context sufficiency. – Benchmarked existing models and established baseline performance metrics for retrieval and generation quality. – Audited and optimized AI systems at Steerlab to improve performance, reliability, and cost efficiency. – Tested new models and optimizations including RAG, fine-tuning, and inference efficiency improvements. – Validated enhancements through controlled experiments before deployment with clear metrics and trade-offs.

– Built a Mac app using AI to auto-rename screenshots and create a fully private, offline search index. – Attracted adoption from professionals at Google and Facebook for organizing daily screenshot workflows. – Featured by Product Hunt on X (Twitter), garnering 157K+ views and 900+ likes, validating strong market demand.

– Crash Test Copy: Stress-test your copywriting before your customers do. Simulates 100+ buyer personas using LLM agents and evolutionary algorithms to iteratively refine and score copy. – Zerodistract: Interactive 3D PDF reader for researchers. – Graphthem: Deep-research engine that reads 10,000 articles and gives one answer. – Introthem: People search engine powered by AI. – Brain Mimic: Turn your thoughts into emails in one click. – EazyRAG: RAG as an API. – ProductSearch: Natural language search engine for software products. – Graha: Collaborate on your website in real-time like Figma. – JustClarify: Customer support page with built-in video recording. – Typelor: Increase your typing speed in a socialized way.

– Got accepted into the Landing Zone program at the world's best startup campus in France to work on my ideas. – Exploring/Researching multiple product ideas with the help of the Station F community. – Attending workshops and seminars on entrepreneurship, innovation, and technology trends. – Networking with fellow entrepreneurs, mentors, and industry experts from diverse backgrounds. – Gaining access to state-of-the-art facilities and resources to prototype and test ideas.

– Developed a novel way to control browsers using a large language model. – The experiment was built using GPT-4o, Meilisearch, Rust, and Next.js. – It is a fully open-source project with 500+ stars on GitHub. – At the time of its release, it was the most stable way to control browsers using a large language model.

– Developed and implemented a ranking algorithm for Grasp, a social search engine. – Attracted early users and generated traction for Grasp. – Leveraged connections to personalize search results based on users' interests. – Received support from renowned investor Elad Gil and co-founder of Twitter Evan Williams. – Featured twice on the front page of HackerNews, validating the idea's significance.

– Contributed to the improvement of document editing and collaboration user experience. – Designed a user experience for mass-importing projects. – Collaborated closely with the development team to ensure the feasibility and implementation of design solutions, iterating based on user feedback and technical constraints. – Conducted user research to gather insights and understand user needs and pain points, resulting in data-driven design decisions.

– Designed and developed a cookie consent banner library that follows the GDPR. Primarily it was created to provide all of the company websites with a legal cookie consent solution. – Contributed to the overall improvement of the company's web development processes by sharing knowledge, conducting code reviews, and suggesting optimizations for performance and usability.