Applied AI/ML Engineer

This is a fully remote position, open to applicants in United States.

📋 Description

• Oversee the complete lifecycle of AI-driven products from initial prototype to full-scale production.

• Choose appropriate models, deploy them, assess their performance, and refine as necessary.

• Implement and enhance LLM inference across the GPU fleet utilizing vLLM and SGLang.

• Optimize continuous batching, KV-cache management, quantization, speculative decoding, and multi-model routing to maximize throughput while minimizing latency and cost per token.

• Develop and manage reinforcement-learning and post-training pipelines using slime and Prime Intellect.

• Design reward systems and verifiers, coordinate rollouts, synchronize weights, and ensure stable, long-term training processes.

• Create evaluation harnesses and benchmarks to measure quality, throughput, and costs.

• Collaborate with Infrastructure on GPU scheduling and fleet optimization.

• Work closely with Product on roadmap priorities.

⛳️ Requirements

• A minimum of 3 years of experience delivering ML/AI systems into production.

• Practical experience with serving LLM inference using vLLM, SGLang, or TensorRT-LLM.

• Familiarity with RL/post-training techniques (GRPO, PPO, DPO, or SFT) or strong related experience coupled with a keen interest in deepening knowledge in this area.

• Proficient in Python and PyTorch.

• Solid understanding of GPU execution, including batching, memory management, and basic CUDA concepts.

• Ability to navigate ambiguity with a strong inclination for proactive action.

• Preferable direct experience with slime, prime-rl, the verifiers library, or Megatron-LM.

• Experience in distributed training methods, including FSDP, TP/PP/DP parallelism is preferred.

• Knowledge of quantization (FP8/INT8), P/D disaggregation, or speculative decoding is a plus.

• Familiarity with verifiable inference or large-scale distributed systems is advantageous.

• Experience with Kubernetes and container-based deployment is favored.

• Understanding of GPU fleet orchestration tools, such as Ray, SkyPilot, or Slurm, is preferred.

• A public GitHub profile must be included in the application.

• The GitHub profile should reflect a minimum of 1 year of activity/history.

• Applications lacking a GitHub profile or showing insufficient activity will not be considered.

🏝️ Benefits

• Equity allocation.

• Comprehensive health, dental, and vision coverage for U.S. employees; regionally adjusted for global employees.

• Flexible Paid Time Off (PTO).

• Budget for professional development and conference travel.

• Remote-first work environment.

• Regular off-site events.

• High-trust, high-velocity team culture.