Skip to content

Work with Trelis

PAID Positions – 1. Model Engineer, 2. Voice Dataset Engineer

Applications are only accepted from individual engineers, not agencies.

Comp: $2k-$3k per week, depending on experience/bandwidth.

Remote contract position for now.

Apply via DM or email voice at trelis dot com.

For the Model Engineer Role, include a) a link to work you have done customising and training model architectures for LLMs or Voice models (e.g. Mamba, global-local attention, TRM, Transformers), b) a link to a project you have coded that you’re proud of.

For the Voice Dataset Engineer Role, include a) a link to work you have done preparing datasets for ASR and/or TTS, b) a link to a project you have coded that you’re proud of.

You’ll work closely with me and clients to launch new model architectures or benchmarks+datasets for voice.


Short Open Source Research Collaborations

I’m starting some short collabs on specific research projects where:

  • – I’ll provide compute, if needed
  • – Work will be done in a public GitHub repo, Apache-2 licensed
  • – This isn’t hiring or paid work

Initial projects:

  • – NanoChat but with a recursive transformer [COMPLETED]
  • – VARC but dropping task embeddings [SKIPPING, IDEA LIKELY DOES NOT WORK AS AUGMENTATIONS “BREAK” CERTAIN TASK DEFINITIONS]
  • – Gather/publish an NVARC-style dataset for ARC-AGI-II
  • – Generate ARC tasks using ASAL from Sakana

If interested, email [email protected] with the specific project + anything you’ve built before (to give a sense of what you’ve worked on).