ML Engineer (Training Infra), Foundational Models
Sarvam AIBengaluruOn-Site
Needs a free Superjobs account. You'll finish on Sarvam AI's Ashby page.
Skill score
Summary
The ML Engineer owns the distributed training infrastructure for Sarvam's next foundational models, covering parallelism, GPU kernel optimization, throughput tuning, and reliability of long training runs. It suits engineers with deep systems and GPU experience who want to keep frontier pretraining runs stable and efficient.
- Function
- Data Science
- Estimated salary
- Est. ₹9.5–14 LPA
- Estimated for a mid-level role, ~3 yrs experience, India. Not published by the employer.
- Work mode and location
- On-Site, Bengaluru
- Experience
- 3+ yrs
- Employment type
- Full-time
Job description
Overview
You will own the infrastructure that our next family of foundational models is trained on. This is a deep systems role: distributed training at scale, parallelism strategies, GPU kernel work, throughput and stability optimization, and the unglamorous reliability engineering that determines whether a frontier training run actually finishes.
The bar is high. A bad MFU number, a slow data loader, a switch flap — these are problems whose cost is measured in weeks and millions of dollars. You will be expected to find them before they happen, and to fix them fast when they do.
Roles and responsibilities
- Build, maintain, and continually push the limits of our distributed training stack across large GPU clusters.
- Design and implement parallelism strategies — data, tensor, pipeline, sequence, expert — and reason about which combinations make sense for which architectures and which scales.
- Profile and optimize end-to-end training throughput: kernel performance, communication overlap, memory layout, checkpointing, data loading.
- Write and tune custom GPU kernels (CUDA, Triton) where off-the-shelf is leaving performance on the table.
- Own the reliability of long-running training jobs — fault tolerance, checkpoint integrity, deterministic restarts, automated detection of slow nodes and silent corruption.
- Partner closely with researchers to make sure architectural ideas can actually be trained efficiently, and with the data team to keep the pipeline from being the bottleneck.
Skills required
Top skills
- Distributed training
- GPU kernels
- CUDA
- Megatron-LM
All skills (8)
- DeepSpeed
- FSDP
- Nsight profiling
- PyTorch internals
- Triton
- NCCL
- NeMo
- Slurm
Qualifications
- BS or MS in Computer Science or a closely related technical field (or equivalent demonstrated experience).
- 3+ years of experience building ML training infrastructure or large-scale distributed systems. Exceptional early-career candidates with a strong systems background will be considered.
- Hands-on experience training large models with distributed training frameworks — Megatron-LM, DeepSpeed, FSDP, NeMo, or equivalent. You should have been on-call for a real pretraining run.
- Deep working knowledge of GPU architecture, the CUDA programming model, and standard tools for profiling GPU workloads (Nsight, PyTorch profiler, etc.).
- Strong PyTorch internals knowledge. You're comfortable reading and modifying low-level training code, not just using high-level APIs.
- Meaningful open-source contributions in the training infrastructure ecosystem — Megatron, DeepSpeed, PyTorch, vLLM, Triton, NCCL, or similar.
Nice to have
- Custom CUDA or Triton kernel development with measurable wins on real workloads.
- Direct experience training models at 10B+ parameters or on 1000+ GPU clusters.
- Familiarity with cluster orchestration and job scheduling (Slurm, Kubernetes) at scale.
- Experience with mixed precision (BF16, FP8), quantization-aware training, or other numerical work that has shown up in production training runs.
- First-author papers or technical reports on training systems, scaling, or model efficiency.
Company info
About Sarvam AI
Sarvam AI is a sovereign AI platform for India, offering speech-to-text, text-to-speech, translation, and conversational agents across 22 Indian languages. Its products are used by enterprises and public institutions.