We develop frontier AI methods to reinvent chip development, collapsing traditional design timelines, closing the loop between AI and the hardware that fuels it, and enabling a Cambrian explosion of custom silicon. Backed by Lightspeed, Sequoia and Felicis.
About the role
Design and iterate on SFT and RL recipes for large-scale models, with a focus on compute efficiency and steep, stable RL climbs. Build and scale synthetic data pipelines, including generation, quality filtering, and coverage metrics. Found on 1752vc Careers, the job board for startup and VC roles.
What they're looking for
- Degree in CS, EE, math, or a related field, or equivalent practical experience
- Hands-on post-training experience (SFT, RL, synthetic data and environments) at a frontier AI lab or equivalent
- Deep fundamentals in transformer internals and distributed training
- Research track record in RL, NLP, or machine learning, such as publications at NeurIPS, ICML, ICLR, or COLM
- Experience building RL environments for complex frontier coding and agentic benchmarks
- A shipped result: a post-training run, data recipe, model release, or research outcome that moved a benchmark or product
More about this role
Ricursive Intelligence is a frontier AI lab building self-improving systems, starting with chip design. We are reinventing chip development and closing the loop between AI and the hardware that fuels it, recursively accelerating the path to artificial superintelligence. Backed by $335M from Sequoia, Lightspeed, DST, and NVIDIA Ventures, we are a growing, fast-paced team where every hire shapes the work.
The company has unmatched talent density, including IMO, IPhO, and IOAA gold medalists, pioneers who made prior breakthroughs in chip design: AlphaChip (Nature 2021), ePlace (DAC Best Paper Nominee 2014), RL-CCD (DAC Best Paper 2023), INSTA (DAC Best Paper 2025), and C3PO (ASP-DAC Best Paper 2026), chip leads for Apple Silicon (multiple generations of iPhone and iPad) and Google (TPU, OpenTitan), and top researchers and engineers from Anthropic, Google DeepMind, Stanford, and MIT.
We are looking for the very best post-training researchers and engineers in the world.
A small team trained Ripple-65B Preview from scratch and set a new U.S. open-weight state-of-the-art on agentic coding (80.5 on TerminalBench 2.1), outperforming U.S. open-weight models up to 15x larger at one-tenth the...
Browse similar: Startup jobs · San Francisco Bay Area