Activities planned for 2025

These are mainly the open source efforts I will be a part of. Most are AI related. The core principles are defined by the AI Punk’s manifesto

1. LLM

   1. Telugu dataset

   2. Research on low resource datasets

   3. Research on mobilizing crowd source dataset collection

   4. Creating a foundational Telugu model

   5. Shape of stories

   6. New sentence embeddings

   7. Chunking

   8. Quantizing/ShorteningLLMs

   9. SLM with synthetic data

2. Speech

   1. Telugu Dataset

   2. Research on low resource dataset for ASR

   3. Transfer learning, from Telugu to all other languages.

   4. Mobile models

   5. Foundational models in speech without using Transformers

   6. TTS dataset

   7. 10,000 TTS voices

   8. Voice cloning

3. Vision

   1. New image embeddings

   2. Image RAG

   3. Image generation using low resources

   4. Image to text

4. Community

   1. Dataset and model licenses

   2. Crowd sourcing templates for the global south

   3. Upskilling template

   4. Upskilling 1 lakh engineering students

   5. Intern repository with ranking

   6. Training tools in English and Telugu

   7. Data collection app and tools

   8. Cluster compute for inference

If anyone wants to take part in any of these activities, please reach out and I will get you involved.