Built a phoneme-level pronunciation scorer ποΈ
Audio + target sentence β per-phoneme scores and a coaching tip for the worst sound.
No fine-tuning, no LLM. Runs 3.9x real-time on one CPU core.
Demo π
#NLProc#SpeechProcessing#MachineLearning
For anyone who wants to understand how Transformers are implemented from scratch in PyTorch, this article is an excellent resource.
https://t.co/bM1uulHOpU
Built a Multi-Agent AI Research System today π
π Web search with Tavily
π§ Research/report generation using Mistral AI
π Source extraction with BeautifulSoup
π€ Optional voice input support
Supports both CLI + local web UI.
GitHub:
https://t.co/BVyy3vTMAA
#AI#AgenticAI