Hey HN,
I built this project because I kinda wanted to build something different from what I usually build. It's a financial sentiment analysis tool using distilled FinancialBERT which was quantized to int8 to run smoothly on a cheap cpu-only VPS without consuming much CPU or Memory.
The Stack:
Inference: ONNX Runtime (Distilled FinancialBERT quantized to Int8).
Backend: Python (Celery) for the heavy lifting + a simple FastAPI app to serve the frontend.
Frontend: SvelteKit (visualizing the live data).
I wrote more about it here:
https://mohammedeabdelaziz.github.io/articles/trendscope-mar...
You can view the repo here:
https://github.com/MohammedEAbdelAziz/TrendScope
I am happy to answer any questions or discuss any thing related to it.
Thanks
