This space is a home for projects funded and built by me, aligned with my vision for applied AI. My focus is on taking the latest advances and turning them into practical systems that solve real-world problems. Applied AI gives us that rare opportunity to create tangible impact, and I feel incredibly lucky to be building in this era.

What I'm Building:

1. PresGenie (https://www.presgenie.in/)

An AI-powered prescription writing tool currently on version 1.1.0. We have good feedback from testers before the official release (Coming Super Soon).

2. NCERTVideoGen - Open Source Video Generation Framework (https://github.com/aditya699/NCERTVideoGen)

I built a production-grade multi-agent system for automated video generation—transforming text prompts into professional Instagram Reels with AI-generated visuals, voiceovers, and sound effects. After 1.5 months of development, I discovered ElevenLabs had launched a similar product. Instead of competing, I'm open-sourcing the entire framework as a learning resource for anyone building async agent systems with multimodal pipelines.

Tech: 4-agent architecture, Redis pub/sub for real-time progress streaming, Celery for async processing, multi-provider fallback (Gemini → OpenAI), FFmpeg audio mixing, MongoDB state management.

Sample video: https://www.youtube.com/watch?v=UCn8yG4hfic

3. EduMOE - AI Research (https://github.com/aditya699/EduMOE)

I have a deep interest in AI research. EduMOE helps people train LLMs from scratch using Mixture of Experts. We have a research roadmap focused on studying LLMs in-depth first, with plans to contribute to the Indic setting soon. Applied research takes time, but hopefully we'll get there.

4. Deep Analysis Agent (https://github.com/aditya699/Deep-Analysis-V1)

We built an agent that does data analysis and creates reports like a junior analyst. The code is open-sourced for anyone to use.

5. Scribe AI (https://github.com/aditya699/scribe-ai)

A WebSocket-based real-time medical transcription and WhatsApp chat service. We didn't continue with it due to poor LLM performance on medical conversations and the high cost of any wrong decisions.