SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Trajectory is building a platform for continual learning in AI systems. As an ML Infrastructure Engineer, you will design and optimize the infrastructure that enables AI systems to improve their own training, inference, and GPU kernels. The team aims to unlock 10× improvements monthly across the stack—from GPU scaling and model size to throughput, memory efficiency, caching, and latency.
You will work across three core areas, with deep expertise in at least one and curiosity across the full stack:
**Training Infrastructure**: Build and optimize distributed training and reinforcement learning infrastructure, including rollout execution, GPU scheduling, checkpointing, and recovery mechanisms. Improve training experimentation throughput and shorten research iteration cycles.
**Inference Systems**: Optimize serving for production agents and training rollouts. Improve batching, scheduling, and KV-cache management while balancing latency, throughput, cost, and model quality.
**Kernels and Runtimes**: Develop and optimize GPU kernels and runtimes using CUDA, Triton, or comparable tools. Improve memory use and execution efficiency while preserving numerical correctness and validating gains in real workloads.
**Cross-functional Work**: Build reproducible benchmarks, observability systems, and automated research workflows that propose infrastructure changes, run experiments, and validate improvements. Collaborate closely with researchers to turn new algorithms into reliable production systems.
Trajectory's core innovation is a continual learning loop for ML infrastructure itself: agents propose optimizations, run experiments, measure gains, and learn from results across training, inference, and kernels.
The company was founded by researchers from DeepMind, OpenAI, Meta Superintelligence, and product leaders from Figma, Apple, Stripe, and Windsurf. Trajectory has raised $60M from Sequoia, Conviction, Jeff Dean, and Fei Fei Li, and is working with customers including Harvey, Rogo, Mercor, Decagon, and Clay.
**Requirements:**
- Strong fundamentals in distributed systems, networking, storage, and failure recovery, with demonstrated experience shipping and operating demanding systems.
- Deep specialization in at least one of: training infrastructure, inference systems, or GPU kernels, supported by systems you have built or measurable optimizations you have delivered.
- Strong Python skills and languages relevant to your specialty (e.g., C++, CUDA, or Triton).
- Understanding of PyTorch, JAX, or comparable framework internals, with strong profiling and debugging skills.
- Ownership mindset and clear communication: ability to work closely with researchers, deliver measurable performance gains, and preserve correctness and reliability.
- Demonstrated capability is valued over credentials.
**Preferred:**
- Hands-on experience with training and RL stacks such as Miles, SkyRL, Prime Intellect's verifiers, or comparable systems.
- Depending on your specialty: experience with vLLM, SGLang, collective communication, or ML compilers.
