Member of Engineering (Multimodality - Research Lead)
Remote
RemoteFull TimeSenior LevelSmall
Full TimeSenior LevelSmall
Job Summary
Own Poolside's multimodality direction and capability adoption by shipping first image-input capabilities and charting the path from adapter-based to native multimodality. Run end-to-end experiments covering hypothesis, implementation, training at scale, and analysis while partnering with evals, architecture, data, and post-training teams to land multimodality in Poolside models. Collaborate on custom evaluations and datasets for multimodal SWE capabilities, drawing on the company's powerful model factory and thousands of GPUs. Build a new team from the ground up alongside founding engineers to shape this frontier capability.
Required Qualifications
- Experience in end-to-end training of production-grade VLMs
- Strong LLM training fundamentals: transformers and distributed training at scale
- Strong Python programming skills
- Pragmatic and high-ownership: you reach for the simplest thing that works and thrive in greenfield ambiguity
Desired Qualifications
- Leadership and 0-1 experience, especially with substantial breadth, ownership, and hands-on contributions
- Distinguished research background on VLMs or multimodal models
- Understanding of the challenges of native multimodality
- Experience building agentic computer use systems
Hiring someone like this?
Get your role in front of qualified candidates on Sorce.