DELHI · FULLTIME
Senior Engineer – Distributed Training Systems & Performance
EQ
Eternity Quests LLP
Delhi · onsite · Posted 39d ago
Sourced from
Not disclosed4–12 yrsfulltimeDelhi
About a minute. You'll make an account on the way through.
agent · working
- Read this posting0.8s
- Found the form on Foundit · 9 questions to answer1.5s
- Decided the answers for all of them2.9s
- Tailor your résumé to this role
- Waiting for you
Section · 01
About this role
- Location: Delhi NCR
- Exp - 4+ years
- Core Mandate: JAX/XLA/Pathways stack, sharding, FP8 numerics, checkpointing, MFU.
- Responsibilities:
- Optimize the core AI training infrastructure utilizing the JAX/XLA and Pathways stack.
- Design efficient tensor, pipeline, and data sharding strategies across massive TPU/GPU clusters.
- Implement FP8 numerics, architect reliable high-speed checkpointing systems, and continuously profile workloads to maximize Model FLOPs Utilization (MFU).
Requirements:
4+ years in distributed systems, High-Performance Computing (HPC), or ML infrastructure. Deep expertise in parallel computing, hardware accelerators, and low-level stack optimization
Sourced from foundit
Section · 02
Skills
GPU optimizationML infrastructureCudaperformance engineeringHpcDistributed SystemsTritonJAXKernel OptimizationNCCLcheckpointingPipeline ParallelismTensor ParallelismXLAModel parallelismFSDPFP8mixed precisionMFU