Remote AI/ML Engineer — Production Inference & RL

Boundless; United States; Hybrid; USD 175,000 - 250,000; Full time

Benefits offered by this job

Competitive salary + equity

Health/dental/vision

Flexible PTO

Dev & conference budget

Remote-first with off-sites

Job summary

Boundless is building toward AI leadership by delivering end-to-end AI features on a growing GPU inference fleet. You will own AI products from prototype to production, optimizing LLM inference, and implementing RL/post-training harnesses to ensure stability and efficiency.

You’ll work with vLLM, SGLang, and Megatron-based tools, focusing on throughput, latency, and cost per token while operating with autonomy in a fast-moving environment.

Qualifications

• 3+ years shipping ML/AI systems to production.
• Experience with LLM inference using vLLM, SGLang, or TensorRT-LLM.
• Experience with RL/post-training methods (GRPO, PPO, DPO, or SFT).
• Strong Python and PyTorch development skills.
• Understanding of GPU execution: batching, memory, CUDA basics.
• Ability to operate with autonomy and bias for action.

Responsibilities

• Own AI features from prototype to production — model selection, serving, evaluation, and iteration.
• Deploy and optimize LLM inference across GPU fleet.
• Build RL and post-training harnesses using slime (Megatron-LM + SGLang) and Prime Intellect (prime-rl + the Environments Hub / verifiers).
• Build eval harnesses and benchmarks measuring quality, throughput, and cost.
• Collaborate with Infrastructure on GPU scheduling and fleet utilization, and with Product on what to build next.

Skills

ML systems production

LLM inference experience

Python

PyTorch

CUDA basics

Ambiguity tolerance

LLM inference stacks

RL / post-training methods

Distributed training

Tools

Kubernetes

Container deployment

Ray

Slurm

Job description

Boundless is coordinating GPU compute at scale and building toward becoming a leader in AI. As an Applied AI/ML Engineer, you'll ship AI-powered products end-to-end on top of our growing GPU inference fleet — owning everything from serving low-latency inference to standing up reinforcement-learning post-training pipelines. This is a builder's role: you take an idea from prototype to production, tune it for throughput and cost on real GPUs, and iterate fast on customer and internal feedback.

You should be comfortable operating with a high degree of autonomy, navigating ambiguity, and defaulting to a strong bias for action.

What You'll Do

End-to-End AI Product Delivery: Own AI features and products from prototype through production — model selection, serving, evaluation, and iteration — shipping working software rather than research artifacts.

Inference Serving: Deploy and optimize LLM inference across the fleet using vLLM and SGLang. Tune continuous batching, KV-cache management, quantization, speculative decoding, and multi-model routing to maximize throughput and minimize latency and cost per token.

RL & Post-Training Harnesses: Build and operate reinforcement-learning and post-training pipelines using slime (Megatron-LM + SGLang) and Prime Intellect (prime-rl + the Environments Hub / verifiers). This includes reward and verifier design, rollout orchestration, weight synchronization, and keeping long-running training stable.

Evaluation & Iteration: Build eval harnesses and benchmarks that measure quality, throughput, and cost together, and use them to drive fast, data-informed iteration.

Work Across the Stack: Partner with Infrastructure on GPU scheduling and fleet utilization, and with Product on what to build next and why.

• 3+ years shipping ML/AI systems to production
• Hands-on experience serving LLM inference with vLLM, SGLang, or TensorRT-LLM
• Experience with RL / post-training methods (GRPO, PPO, DPO, or SFT), or strong adjacent experience and a clear desire to go deep here
• Strong Python and PyTorch
• Working understanding of GPU execution: batching, memory, and basic CUDA concepts
• Comfort operating in ambiguity with a strong bias for action

Nice to Have

• Direct experience with slime, prime-rl, the verifiers library, or Megatron-LM
• Distributed training experience (FSDP, TP/PP/DP parallelism)
• Quantization (FP8/INT8), P/D disaggregation, or speculative decoding
• Experience with verifiable inference or large-scale distributed systems
• Kubernetes and container-based deployment
• Familiarity with GPU fleet orchestration (Ray, SkyPilot, Slurm)

Additional Requirements

• Candidates must include a public GitHub profile in their application.
• The GitHub profile should demonstrate a minimum of 1 year of activity/history.
• Applications that do not include a GitHub profile, or show insufficient activity, will not be considered.

At Boundless, we take care of our people, because building the future of AI compute starts with an empowered team. Here's what you can expect when you join us:

• Competitive salary (proposed band b/t US$175k and $250k annually) + equity allocation
• Health, dental, vision (for U.S. employees; region-adjusted globally)
• Flexible PTO
• Professional development and conference travel budget
• Remote-first with regular off-sites and a high-trust, high-velocity team environment

We are a global team, and applicants from around the world are welcome to apply.