Founding ML Researcher

Tykhe Inc; San Francisco (CA); On-site; USD 120,000 - 210,000; Full time

Job summary

Tykhe Inc. is seeking a producer of VLM and multimodal AI systems for on-site work in San Francisco. You will ship production-grade visual-language models, orchestrate end-to-end workflows, and contribute to tool-use driven agentic AI projects.

Ideal candidates have 0–3 years of experience, strong Python/PyTorch skills, and a background in vision or multimodal research. A startup or high-intensity environment is highly valued.

Qualifications

• Shipped production VLM / multimodal AI systems.
• Experience with Python and PyTorch.
• On-site in San Francisco 5 days/week.
• BS or MS in CS/ML/Engineering or equivalent via shipped production AI systems.
• Vision-language multimodal focus; agentic AI with tool use.
• Startup or high-intensity company experience preferred.

Skills

VLM engineering

Multimodal AI

Eval design

Model orchestration

Python programming

Education

BS or MS in CS/ML/Engineering

MS with vision or multimodal research component

Tools

Python

PyTorch

vLLM

Triton

TensorRT

Job description

Education

Shipped agentic VLM / multimodal vision systems to real users in production — not demos, not pure research. Owned the model + orchestration layer end- to- end. Must be vision- language multimodality (images/video + VLM), not sensor- fusion or audio- only.

Applied VLM engineering: shipped and hardened vision- language models in production — visual reasoning, detection/segmentation as needed. Depth in applied fine- tuning (SFT, RLHF), eval design, and model orchestration, not pretraining from scratch.

Must work on- site 5 days/week in San Francisco. Ideally open to living in a hacker house.

0 - 3 years of experience shipping production VLM / multimodal AI systems, using Python and PyTorch

Startup or founding- engineer experience (founder, CTO, eng #1–5) OR high- intensity company (Meta Reality Labs, Anduril, SpaceX, top YC). Long single big- tech tenure with no builder signal and no relevant- team work is a negative.

Education

BS or MS in CS, ML, or Engineering from a strong program, OR demonstrated equivalent via shipped production AI systems.

Agentic AI with real tool- use: built multi- step agent loops (not RAG wrappers or single- turn chatbots). Experience with eval harnesses and data flywheels.

Genuinely mission- aligned with computer vision, wearables, and industrial AI — not just chasing LLM trends. Evidenced by project choices, side work, or career trajectory.

Nice-to-have

Production AR / wearable AI (Meta Reality Labs, Snap Spectacles, Apple Vision Pro) or autonomous driving CV (Waymo, Cruise, Aurora).

Education

Master's with vision or multimodal research component (published work or thesis).

Edge inference / on- prem model deployment (vLLM, Triton, TensorRT, quantization). Serving open- weight models on constrained hardware.