ML Engineer (Training Infra), Foundational Models

Sarvam; Bengaluru; On-site; INR 3,500,000 - 6,000,000; Full time

Job summary

Sarvam is seeking a skilled ML infrastructure engineer to own the training platform for large models. This deep systems role covers distributed training at scale, GPU kernel work, and reliability engineering to ensure frontier runs finish on time.

You will design parallelism strategies, optimize throughput, and tune CUDA/Triton kernels. Collaborate with researchers to ensure architectures train efficiently at scale.

Qualifications

• BS or MS in Computer Science or equivalent.
• 3+ years of experience in ML training infrastructure or large-scale distributed systems.
• Hands-on experience with Megatron-LM, DeepSpeed, FSDP, NeMo, or equivalent; on-call for real pretraining runs.
• Deep knowledge of GPU architecture, the CUDA programming model, and profiling tools (Nsight, PyTorch profiler).
• Strong PyTorch internals knowledge; able to read/modify low-level training code.

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 scale.
• 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, slow-node detection, corruption checks.
• Partner closely with researchers to ensure architectural ideas train efficiently and with the data team to avoid bottlenecks.

Skills

Distributed training

GPU kernel programming

PyTorch internals

On-call for pretraining

Performance profiling

Education

BS/MS in Computer Science

Tools

Megatron-LM

DeepSpeed

FSDP

NeMo

CUDA

NVIDIA Nsight

Kubernetes

Slurm

Job description

About Sarvam

Sarvam is building the bedrock of Sovereign AI for India. The company is developing India’s full-stack sovereign AI platform, building across research, models, infrastructure and applications with a singular focus on making AI genuinely work for India. Sarvam works with leading enterprises and public institutions and is backed by Lightspeed, Peak XV, and Khosla Ventures. Sarvam partners with India’s leading brands, including Tata Capital, SBI Life, CRED, IDFC, and LIC.

About the Role

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.

What You’ll Do

• 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.

What We’re Looking For

• 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.

Bonus Points

• 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.

Why this role?

Frontier model training is one of the hardest systems problems in the world right now, and most of the people doing it well sit at a small handful of labs. You will be doing the same work, with the same autonomy, and your impact will be visible in every model Sarvam ships.

Why Sarvam?

Sarvam is a fast-moving, high talent-density team building full-stack AI for India, working on problems that push the frontiers of AI with real population-scale impact.

• Work alongside researchers, engineers, builders, and business leaders who move fast and hold each other to a very high bar
• High ownership and high impact, from day one
• Everything we do is AI-first, from the way we build and ship to the way we think about problems
• You can work on problems that could change how an entire country learns, works, and communicates

If you want to work on problems at the frontier of AI in India, Sarvam is the place to be.