Audio ML Noise Reduction
2026Real-time background noise suppression for Teams and Zoom. Custom U-Net trained on DNS Challenge 4. ONNX CPU inference. Built from scratch — dataset, model, training, live pipeline.
View repo →Stavion
Colquitt
Audio ML Engineer, Senior AV/AI Engineer, and Recording artist in Chicago. I build stuff!
10+
Open-source
projects
projects
5+
Records
released
released
∞
Domains
still to learn
still to learn
01
Work
Real-time speech enhancement, edge ML, and production AI architecture and much more! Selected open source below — more on GitHub.
Hybrid RAG Pi
2026Production hybrid retrieval combining Qdrant (vector) and SQLite FTS5 (keyword), plus hierarchical conversation memory with LLM-powered fact extraction. Deployed on Raspberry Pi 5.
View repo →Audio Analysis MCP
2026Real-time audio analysis as an MCP server — frequency reports, stereo imaging, mix analysis, key detection. Built to let AI assistants reason about what's playing during a production session.
View repo →Spectrogram Reader
2026See what audio processing actually does. Interactive STFT explorer across linear, mel and MFCC scales — plus an A/B difference view on a shared dB reference that shows exactly what a plugin, filter or bounce changed.
View repo →
02
Music
I write and produce as Stavion Colquitt. Singer, pianist, and engineer of my own records. New material in the pipeline; older work below.
03
Research
Write-up in progress
Current project: an AI backing vocalist. Existing tools execute an arrangement decision you have already made — pick the harmony, set the interval, print it. This system makes the decision instead. Given a lead vocal, the instrumental and the lyrics, it predicts what belongs on each beat: a double, a harmony, a call and response, an ad‑lib, a sound effect, or nothing at all.
Trained on separated stems from my own catalogue, where I know what was sung and why it was sung there. Being written up with two research advisors. More here once it is further along.
04
Study
-
2026–2028
M.S., Artificial Intelligence
Lewis University
Currently enrolled -
2025
CCNP Collaboration
Cisco
-
2025
Dante Level 3
Audinate
-
2022
B.S., Digital Audio Recording Arts
University of St. Francis
06
Connect
Best for: research collaborations, audio ML conversations, music questions. Slow for cold outreach — I read carefully before I reply. For more on who I am, head to about me.
Socials
Where I show up online — music drops, work in progress, the occasional life moment.