Strategic AI Observatory
ML learning and agent development — the workshop behind Randolph Mbecheun's exploration of applied intelligence.
Hands-on machine learning projects — built end-to-end, from CLI prototype to deployment. Audio and generative AI are the current focus.
End-to-end noise reduction for audio and video files using DeepFilterNet. CLI auto-detects format, extracts the audio stream with FFmpeg, denoises it, then muxes it back into the original video without re-encoding the picture. Runs at ~9× realtime on CPU. Serverless AWS deployment (S3 + Lambda + DynamoDB) in progress.
AI-powered text-to-speech with personality. Generate Mode turns a script into a WAV across five distinct vibes — Professional, Silly, Chaos, Motivation, Tech Storyteller. Theater Mode renders multi-character dialogue, giving every speaker their own voice and tone.
Building autonomous and semi-autonomous agents — exploring how language models can plan, act, and coordinate with real systems.
A spreadsheet you edit by talking to it. Upload an Excel or CSV file, type an instruction in plain English — sort by date, add a calculated column, drop empty rows — and the editor rewrites the grid for you. Export the result back to .xlsx. The model runs server-side; no API keys touch the browser.