Bitdeer Technologies Group is seeking a Senior AI Scheduling & Orchestration Engineer to join their AI cloud division. The role focuses on designing and implementing advanced batch scheduling architectures to support high-volume AI workloads in large-scale HPC environments. Responsibilities include: designing multi-node gang scheduling using frameworks like Volcano or YuniKorn; managing cluster-wide admission control and job queueing with Kueue; leveraging Kubernetes Dynamic Resource Allocation and custom scheduler plugins; architecting topology-aware pod placement for low-latency communication via NVLink and InfiniBand; and implementing GPU sharing technologies like MIG and time-slicing. You will collaborate with GPU Systems and Storage teams to integrate the scheduling layer with bare-metal hardware and mentor junior engineers. Requirements: Bachelor’s or Master’s degree in Computer Science, Electrical Engineering, or related field; 6+ years of distributed systems engineering with deep expertise in Kubernetes scheduling; experience with AI workload execution patterns and distributed training frameworks (PyTorch Distributed, Ray, MPI); proven track record in scaling scheduling stacks in HPC or cloud environments; strong knowledge of GPU hardware architectures; and experience with infrastructure automation (Terraform, Go-based Operators). The company offers an inclusive, fast-growing environment with opportunities for professional development and impact on the digital asset and AI cloud industry.
For full details about this role, click “Apply on company site” to view the employer’s original posting.
A focused preparation guide, built around this job’s responsibilities and requirements.
✦ AI-generated guide
Preparation suggestions, not the employer’s actual interview questions. Always check the original posting for current requirements.
01
Understand the role
What matters most
Advanced Batch Scheduling Architecture
Design and implement multi-node gang scheduling using frameworks like Volcano or YuniKorn to support high-volume AI workloads, ensuring efficient resource allocation in large-scale HPC environments.
Topology-Aware Resource Orchestration
Architect pod placement strategies that account for hardware topology to minimize latency across NVLink and InfiniBand fabrics, directly impacting the performance of distributed training frameworks like PyTorch and Ray.
GPU Utilization and Multi-Tenancy
Implement and manage GPU sharing technologies such as MIG and time-slicing to maximize hardware utilization while maintaining strict multi-tenancy isolation policies.
Infrastructure Reliability and Automation
Develop and maintain the scheduling stack using Go-based Operators and Terraform, focusing on resolving resource contention and deadlocks to ensure high availability for AI cloud services.
Likely challenges to think through
Balancing Throughput with Latency
The trade-off between maximizing cluster utilization via batch scheduling and minimizing communication latency for distributed training. Prepare by discussing how you have tuned scheduler plugins to prioritize locality without causing job starvation.
Integrating Scheduling with Bare-Metal Hardware
The difficulty of bridging high-level Kubernetes abstractions with low-level hardware constraints (storage I/O and GPU interconnects). Prepare by reviewing your experience in troubleshooting hardware-software interface issues in HPC environments.
Managing Complex Admission Control
Handling high-volume, heterogeneous AI job queues using Kueue while preventing system-wide deadlocks. Prepare by articulating your approach to designing robust queueing logic that handles varying priority levels and resource requirements.
03
Practice your answers
Technical practice
Implementing Multi-Node Gang Scheduling
What to coverExplain how you would use frameworks like Volcano or YuniKorn to handle gang scheduling for distributed AI training. Discuss the challenges of ensuring all pods in a job start simultaneously to prevent deadlocks and resource fragmentation.
Optimizing Topology-Aware Pod Placement
What to coverDescribe your approach to scheduling pods to minimize latency in environments using NVLink and InfiniBand. Address how you account for physical hardware topology when making placement decisions in Kubernetes.
Managing GPU Sharing and Isolation
What to coverCompare and contrast MIG (Multi-Instance GPU) and time-slicing for GPU sharing. Discuss how you would implement multi-tenancy isolation policies to ensure security and performance stability for different users.
Scaling Job Queueing with Kueue
What to coverHow would you design a cluster-wide admission control system using Kueue to manage high-volume AI workloads? Focus on how you handle resource quotas and priority preemption in a multi-tenant HPC environment.
Integrating Kubernetes with Bare-Metal Hardware
What to coverDiscuss the complexities of integrating a Kubernetes scheduling layer with underlying bare-metal GPU and storage systems. Suggest how you would handle storage I/O patterns to prevent bottlenecks during large-scale distributed training.
Developing Custom Kubernetes Scheduler Plugins
What to coverDescribe a scenario where you would need to build a custom scheduler plugin rather than using standard Kubernetes scheduling. Explain the process of extending the scheduler and the potential impact on cluster performance.
Debugging Distributed Training Frameworks
What to coverHow do you troubleshoot performance issues in distributed training jobs running on PyTorch Distributed, Ray, or MPI? Focus on identifying whether the bottleneck is in the network, storage, or GPU scheduling.
Infrastructure Automation with Go Operators
What to coverExplain the role of Go-based Kubernetes Operators in managing AI infrastructure. Provide an example of how you would automate the lifecycle of a complex resource using a custom controller.
Behavioural practice
Mentoring Junior Engineers
What to coverDescribe a time you mentored a junior engineer on a complex technical project. Focus on how you balanced providing guidance with allowing them the autonomy to learn and solve problems independently.
Resolving Cross-Functional Conflicts
What to coverTell me about a time you had to align different teams, such as GPU Systems and Storage, on a shared architectural goal. How did you manage competing priorities and technical disagreements?
Handling Ambiguous Requirements
What to coverGive an example of a project where the requirements were initially ambiguous. How did you translate those needs into a concrete, scalable engineering solution?
Managing High-Pressure Production Issues
What to coverDescribe a situation where you had to resolve a critical resource contention or deadlock in a large-scale production environment. What was your process for diagnosing the root cause and implementing a fix?
02
Sharpen your knowledge
Topics to prepare
Advanced Batch Scheduling Architectures
Focus on multi-node gang scheduling patterns using frameworks like Volcano or YuniKorn. Exercise: Design a conceptual workflow for a distributed training job that requires simultaneous allocation of 16 GPUs across 4 nodes, ensuring all pods start together to prevent deadlocks.
Topology-Aware Pod Placement
Understand how to optimize pod placement for low-latency communication over NVLink and InfiniBand fabrics. Exercise: Research how Kubernetes scheduler plugins can be extended to account for physical rack and switch topology to minimize inter-node latency for large-scale AI training.
GPU Resource Management and Isolation
Study Multi-Instance GPU (MIG) and time-slicing techniques to maximize cluster utilization. Exercise: Compare the trade-offs between MIG and time-slicing regarding performance isolation and memory overhead for multi-tenant AI inference workloads.
Skills to revise
Kubernetes Custom Scheduler Plugins
Develop proficiency in writing custom scheduler plugins to handle complex accelerator requests. Exercise: Create a simple Go-based scheduler plugin that filters nodes based on specific GPU memory availability or hardware features.
Kubernetes Operators with Go
Implement infrastructure automation using Go-based Operators. Exercise: Build a basic Kubernetes Operator that manages the lifecycle of a custom resource representing an AI training job, ensuring it integrates with Kueue for queueing.
Distributed Training Framework Integration
Gain hands-on experience with PyTorch Distributed, Ray, or MPI execution patterns. Exercise: Configure a Ray cluster on Kubernetes to run a distributed training job, focusing on how the scheduler handles resource requests for the head and worker nodes.
04
Your preparation checklist
Prepare a technical case study on Kubernetes batch scheduling
Draft a summary of a past project where you implemented gang scheduling using Volcano or YuniKorn. Focus on how you handled multi-node job dependencies, resource contention, and how you ensured high-volume throughput in an HPC or cloud environment.
Develop a strategy for topology-aware GPU orchestration
Be prepared to discuss how you would architect pod placement to minimize latency across NVLink and InfiniBand fabrics. Explain your approach to integrating Kubernetes Dynamic Resource Allocation (DRA) with custom scheduler plugins to optimize hardware-aware workloads.
Review GPU sharing and multi-tenancy implementation patterns
Prepare to explain your experience with MIG (Multi-Instance GPU) and time-slicing. Detail how you have implemented these technologies to balance resource isolation with high utilization in a multi-tenant AI training environment.
Synthesize your experience with distributed training frameworks
Create a brief overview of how you have tuned scheduling for PyTorch Distributed, Ray, or MPI workloads. Highlight specific challenges you solved regarding job queueing, admission control with Kueue, and storage I/O integration.
Formulate questions regarding infrastructure automation and team collaboration
Prepare questions about their current Go-based Operator development lifecycle and Terraform workflows. Ask how the scheduling team collaborates with GPU Systems and Storage teams to resolve hardware-level bottlenecks.
Availability last checked: 18 Oct 2026. Status can change after a check; confirm on the employer’s website. Report an issue · Our checking policy
Share this opportunityKnow someone who would be a great fit?