Skip to content

Careers · Research

Member of Technical Staff, Pretraining

Train the vikasit model family — from data to a frontier MoE — at the lowest cost per unit of quality.

ResearchPuneFull-time

About the role

You'll work at the core of LambdaQ: training the foundation models that power every Vikasit product. From data curation and tokenization to large-scale distributed training and post-training, you'll push our models up the frontier while keeping inference economics sane.

What you'll do

  • Own parts of the pretraining pipeline — data, architecture, or training infrastructure
  • Run and analyze large-scale training runs on multi-node GPU clusters
  • Improve data quality, mixtures, and curricula for Indian and global use
  • Design experiments that move benchmark and downstream quality measurably

What we're looking for

  • Strong PyTorch and distributed-training experience (FSDP / DeepSpeed / Megatron)
  • Solid grasp of transformer internals, optimization, and scaling laws
  • Experience training models at 1B+ scale, or equivalent research depth

Nice to have

  • MoE training experience
  • Tokenizer / data-pipeline work
  • Publications at top ML venues

Sound like you?

We hire for skill over credentials. Tell us why you're a fit — links and projects welcome.

Apply for this role