Applied Research - Evals & Data

Prime Intellect•San Francisco, CA

71d•Remote

About The Position

Building Open Superintelligence Infrastructure Prime Intellect is building the open superintelligence stack - from frontier agentic models to the infra that enables anyone to create, train, and deploy them. We aggregate and orchestrate global compute into a single control plane and pair it with the full RL post-training stack: environments, secure sandboxes, verifiable evals, and our async RL trainer. We enable researchers, startups, and enterprises to run end-to-end reinforcement learning at frontier scale, adapting models to real tools, workflows, and deployment contexts. We recently raised $15mm in funding (total of $20mm raised) led by Founders Fund, with participation from Menlo Ventures and prominent angels including Andrej Karpathy (Eureka AI, Tesla, OpenAI), Tri Dao (Chief Scientific Officer of Together AI), Dylan Patel (SemiAnalysis), Clem Delangue (Huggingface), Emad Mostaque (Stability AI) and many others. Role Impact This is a customer facing role at the intersection of cutting-edge RL/post-training methods, applied data, and agent systems. You’ll have a direct impact on shaping how advanced models are aligned, evaluated, deployed, and used in the real world by: Advancing Agent Capabilities: Designing and iterating on next-generation AI agents that tackle real workloads—workflow automation, reasoning-intensive tasks, and decision-making at scale. Working with applied data from real deployments to continuously refine policies, improve reasoning, and enhance reliability and safety. Building Robust Infrastructure: Developing the distributed systems, evaluation pipelines, and coordination frameworks that enable these agents to operate reliably, efficiently, and at massive scale. Building data capture, processing, and versioning workflows for feedback, model traces, and reward signals. Bridge Between Customers & Research: Translating customer needs and insights from applied data into clear technical requirements that guide product and research priorities. Collaborating closely with RL and eval teams to ensure real-world signals inform model alignment and reward shaping. Prototype in the Field: Rapidly designing and deploying agents, evals, and harnesses alongside customers to validate solutions. Using applied evaluation data to iterate on model performance and discover new capabilities.

Requirements

Strong background in machine learning engineering, with experience in post-training, RL, or large-scale model alignment.
Experience with applied data workflows and evaluation frameworks for large models or agents (e.g., SWE-Bench, HELM, EvalFlow, internal eval pipelines).
Deep expertise in distributed training/inference frameworks (e.g., vLLM, sglang, Ray, Accelerate).
Experience deploying containerized systems at scale (Docker, Kubernetes, Terraform).
Track record of research contributions (publications, open-source contributions, benchmarks) in ML/RL.
Passion for advancing the state-of-the-art in reasoning, measurement, and building practical, agentic AI systems.

Responsibilities

Work side-by-side with customers to deeply understand workflows, data sources, and bottlenecks.
Prototype agents, data pipelines, and eval harnesses tailored to real use cases, then hand off hardened systems to core teams.
Translate customer insights and evaluation results into roadmap and research direction.
Design and implement novel RL and post-training methods (RLHF, RLVR, GRPO, etc.) to align large models with domain-specific tasks.
Build evaluation harnesses and verifiers to measure reasoning, robustness, and agentic behavior in real-world workflows.
Integrate applied data collection and analytics into the post-training process to surface regressions, emergent skills, and alignment opportunities.
Prototype multi-agent and memory-augmented systems to expand capabilities for customer-facing solutions.
Rapidly prototype and iterate on AI agents for automation, workflow orchestration, and decision-making.
Extend and integrate with agent frameworks to support evolving feature requests and performance requirements.
Architect and maintain distributed training and inference pipelines, ensuring scalability and cost efficiency.
Develop observability and monitoring (Prometheus, Grafana, tracing) to ensure reliability and performance in production deployments.