#Python

Building the Video Translator , the assumption going in was that OpenAI Whisper’s larger model would transcribe more accurately. In practice, on clips with loud background music, the turbo model did noticeably better than large — it seems to tolerate messy audio better rather than just being a faster, lower-quality option. Separating vocals from background music first (with demucs) didn’t help as much as expected either; Whisper handled the mixed track better on its own. Worth trying turbo first rather than assuming bigger is more accurate.

A patched fork of OpenVoice that clones a speaker's voice for short subtitle lines and runs on CPU — the TTS stage of the Video Translator pipeline.

A local AI dubbing pipeline that transcribes a video's audio, translates it, and re-synthesizes speech in the target language using open-source models.

Relaxed the minimum-segment-length check in se_extractor.py so sub-1.5s lines can be cloned, and made the pipeline run on CPU (Apple M1).

Blockchain integrations and fault-tolerant wallet-sync workflows for a crypto portfolio and tax platform.

Designed the database architecture for an in-house Global Distribution System and a serverless bus-ticket booking engine.