Kelechi (tensor)

I am an avid machine learning engineer/researcher.

currently working on generative image/audio models.

Skills

Proficient in JAX/Pytorch, using/*bending* open source models.

I usually spend time tinkering and experimenting with stuff.
My main and always-updated portfolio is my Github.

projects(scratch)

  • ripple [text-image semantic search and tagging with CLIP]
  • shira [neural audio search/retrieval based on CLAP]
  • mini_pixart [minimal pytorch implementation of Pixart text2image model, pytorch]
  • aaliyah [mobilenet/convnext vision impls., reusable for image classification]
  • lila [vision transformer/masked autoencoder in pytorch]
  • shiryoku [vision model for image captioning, CNN+RNN]
  • mini_clip [minimal implementation of the CLIP paper]
  • soundnet [music/audio classification, pytorch]
  • paint_diffuse [a mini-diffusion model for unconditional image(painting) generation]
  • ..and many others on my github.

Other cool stuff you can check out

  • play_jax [mini JAX experiments and implementations]
  • moonvid [hacky video captioning with moondream VLM]
  • web_chat [RAG using websites as data source]
  • studyassist [pdf/document chatbot]
  • vidtext [video/audio transcription using Moonshine, w/files and Youtube urls]
  • ...and many more coming :)

Articles

ML/AI research is like crack to me

Well, I don't usually write a lot, but once in a while :)