2025 08 15 dwani gpu grant local
Open Source Grant Proposal – Euro 5000 for RTX 5090 Desktop
Request
We request RTX 5090 - 32GB VRAM Desktop for ongoing open-source development and deployment of dwani.ai, our multimodal inference platform.
During IndiaFoss conference, We got requests from NGO's, Educational institutions to provide Indian language based AI for multiple use-cases.
Proposed Work
-
Build Discovery - Document Analytics platform for Indian languages
- Windows + Linux App - built with Electron to use dwani.ai in Labs/ Offices
- Source Code
- https://github.com/dwani-ai/dwani-desktop
- https://github.com/dwani-ai/discovery
-
Provide inference from Desktop via cloudflare tunnels for Mobile App and Workshop Demo's
- We have experimented with RTX 4050 laptop with 6GB VRAM, but quantized models and small context length greatly affect the accuracy
- https://github.com/slabstech/llm-recipes/blob/main/tutorials/cloudflare/plan.md
GPU Allocation Plan
- Purpose:
- Run Gemma3‑12B‑Instruct with vLLM for large-scale document, image, and text inference.
- Run ASR on CPU for Speech to Text
- Goal: Deliver fast and efficient LLM inference for Indian languages.
Cost Breakdown
Resource | Total (EURO) |
---|---|
1× RTX 5090 | 5,000 |
- Online Quotes
- Euro 4699 - https://www.notebooksbilliger.de/acer+predator+orion+7000+pc+gaming+desktop+871011
- Euro 4299 - https://www.notebooksbilliger.de/asus+rog+g700+mt+gaming+pc+886787
-
Euro 5299 - https://www.notebooksbilliger.de/msi+mpg+infinite+x3+ai+2nvz9+032at+desktop+871164 Grant Budget: $4,500
-
Remaining Balance
- Would be used to provide Free Monthly Workshops to students
Impact
This grant will allow us to:
- Continue multimodal inference without interruption
- Keep API access free for workshops and student developers