Teams at Goaly

Own the research-engineering loop for turning base models into capable, reliable agents. Design RL environments, training and evaluation data, post-training experiments, reward systems, and regression suites. Analyze model behavior, diagnose reward hacking and other failures, improve distributed training and rollout infrastructure, and translate successful research into reproducible production pipelines. The role requires strong Python and software-engineering skills, modern language-model experience, deep-learning knowledge, and reinforcement-learning intuition.

Own the lifecycle of distributed accelerator clusters, including provisioning, upgrades, networking, security, observability, recovery, and decommissioning. Build control-plane services and infrastructure automation, improve fleet reliability and scalability, establish operational metrics and SLOs, and lead incident response. Partner with cloud, hardware, networking, storage, security, and AI engineering teams to operate dependable infrastructure for training, inference, and other communication-intensive workloads.

Build and optimize foundational AI infrastructure for large-scale training, reinforcement learning, rollout, and inference workloads. Responsibilities include profiling bottlenecks, optimizing GPU and distributed execution, developing scheduling and load-balancing systems, modeling performance and capacity, improving fault tolerance, and creating benchmarks and observability tools. The role partners closely with AI researchers and systems engineers to improve throughput, reliability, utilization, latency, and cost across complex GPU clusters.