We had collected thousands of hours of speech data, we needed to transcribe the speech to text. Teams were waiting for GPU credits etc. I was talking to the team at an internship meet where there were some 300 interns. I saw that each intern was carrying a laptop. There was compute sitting in front of us. I told the Viswam team why cant we use the compute of the laptops to transcribe the audio or convert pdfs to text using OCR. If we have 30,000 interns and each student contributes 1 hour of compute, we have 30,000 hours per day :) Viswam team created simple tools in a day to access the data and assign it based on compute donated and do the transcription. Below is the leaderboard till yesterday. Community FTW :)

While I have been busy with some things, the teams at Viswam have been rocking. The community driven effort is going full throttle. Data set contributions across the Telugu states so far. We will be making some announcements next week.

What a high. Finished the AI days 2025. Thanks to Tech Mahindra and @anandmahindra for providing an awesome venue at Hyderabad. Hyderabad is now the place for AI. 1500 participants, 4 closed door round tables, community Telugu LLM promised, society involved, mission accomplished.

If you have to learn how to fine tune LLMs for your use case, you have the best master teaching how at the conference. Either get a good rank and get an admission at IIIT-H and learn or just attend the conference and learn. There can't be a better deal :)