About the roleAs a
Research Scientist - Reinforcement Learning at Percepta, you will work at the intersection of RL research and real-world deployment. You will advance the frontier of capabilities through research on decision-making for critical industries. You will collaborate closely with our Embedded Product Managers (EPMs) and engineers to ensure that our solutions transform how companies operate.
Responsibilities- Identifying which real-world challenges are tractable for RL-guided decision making.
- Develop RL methods to perform complex tasks in domains like planning, decision-making, or optimization.
- Develop and maintain the experimental infrastructure that powers our research, from simulation environments and data pipelines to training and evaluation frameworks.
- Conduct in-the-wild evaluations at scale that drive millions of dollars in value.
- Partner with our applied AI engineers to transition successful research ideas into robust features of our Mosaic platform.
- Communicate research outcomes to both technical and non-technical stakeholders, making sure everyone understands the "so what" of research and how to apply it.
You may be a good fit if you:- Have an MS/PhD in Computer Science, ML, or related field, or equivalent experience.
- Have a track record of effective RL work.
- Are motivated by impact in critical industries including healthcare, supply chains, energy, and finance.
- Understand how to perform rigorous RL experimentation.
- Enjoy extreme ownership.
- Believe that AI can drive transformative change in critical industries.
The following list can be a sign that you might be a good technical fit:
- High performance, large scale distributed systems.
- Large scale LLM training or RL training.
- Possess strong programming skills, especially in Python.
- Implementing LLM post-training algorithms.
- Experience with vLLM/SGLang, Ray, Kubernetes (or AWS EKS).
- Experience with distributed checkpointing, multi-node, multi-gpu training, custom KV-caching.
- Experience with asynchronous training and inference, either with VeRL, ROLL, SkyRL, AReal, or with RL libraries like CleanRL.
We're working against an incredibly ambitious mission. It won't be easy, but it will likely be the most fulfilling work of your career. If this excites you, let's chat, even if you don't meet all of the qualifications above.
