- Design and operate GPU and accelerator infrastructure for training and inference, spanning on-prem clusters, cloud-managed services, and hybrid configurations.
- Build scheduling, queueing, and resource-sharing systems that maximize accelerator utilization across many teams.
- Integrate frameworks such as PyTorch, JAX, DeepSpeed, FSDP, Megatron-LM, and Ray Train into a unified platform offering.
- Operate high-performance storage systems and data pipelines that keep accelerators fed with training data at near-line-rate.
- Design networking architectures supporting RDMA, InfiniBand, NCCL, and high-bandwidth collective communication.
- Build observability for AI workloads including utilization, throughput, training stability, and failure-mode analytics.
- Implement checkpointing, restart, and fault-tolerance patterns for long-running training jobs at scale.
- Drive cost optimization across compute, storage, and networking through scheduling, spot capacity, and right-sizing.
- Develop developer tooling and paved-road workflows that let researchers launch experiments safely and efficiently.
- Partner with research and applied ML teams to plan capacity for upcoming training runs.
- Implement security controls, isolation, and access management for multi-tenant AI infrastructure.
- Drive automation across cluster provisioning, lifecycle management, and configuration enforcement.
- Maintain runbooks, capacity dashboards, and operational documentation for the AI platform.
- Stay current with AI infrastructure research, accelerator hardware, and emerging open-source AI tooling.
- Bachelor’s or Master’s degree in Computer Science or a related field.
- Six or more years of experience in infrastructure, platform, or HPC engineering.
- Hands-on experience operating GPU clusters or large-scale ML training infrastructure.
- Strong proficiency in Python and at least one systems language such as Go or C++.
- Deep understanding of distributed training, accelerator architectures, and collective communication.
- Experience with Kubernetes, Slurm, Ray, or similar scheduling systems for ML workloads.
- Strong understanding of Linux internals, networking, and high-performance storage.
- Experience with at least one major cloud provider’s ML infrastructure offerings.
- Strong software engineering practices including testing, CI/CD, and code review.
- Excellent communication and cross-functional collaboration skills.
- Experience operating InfiniBand or RDMA networking at scale.
- Contributions to open-source ML infrastructure projects.
- Familiarity with custom orchestrators or research-grade training stacks.
- Exposure to frontier model training operations.
- Experience with FinOps for AI workloads.
Equal Employment Opportunity (EEO) Statement
Bright Vision Technologies (BV Teck) is committed to equal employment opportunity (EEO) for all employees and applicants without regard to race, color, religion, sex, sexual orientation, gender identity or expression, national origin, age, genetic information, disability, veteran status, or any other protected status as defined by applicable federal, state, or local laws. This commitment extends to all aspects of employment, including recruitment, hiring, training, compensation, promotion, transfer, leaves of absence, termination, layoffs, and recall.
BV Teck expressly prohibits any form of workplace harassment or discrimination. Any improper interference with employees’ ability to perform their job duties may result in disciplinary action up to and including termination of employment.
Originally posted on Himalayas
