Senior Data Scientist - Gurugram
Core
Build and improve open-weight LLMs (Llama, Qwen, Gemma) for an AI personal assistant serving millions of Indians via phone and WhatsApp.
Role type
Senior IC machine-learning engineer (LLM training & RL)
Builds
AI personal assistant models and training infrastructure
Domain
Consumer AI / Large Language Models
Deliverable
production ML models
Required skills
Open-weight LLM post-training (SFT, DPO, RL), PyTorch, distributed training (FSDP, DeepSpeed), LoRA/PEFT, high-throughput inference (vLLM, SGLang), dataset engineering for AI agents, transformer architecture expertise
Preferred skills
Experience in early-stage/founding teams, product 0-to-1 experience
Technologies
PyTorch, FSDP, DeepSpeed, vLLM, SGLang, Llama, Qwen, Gemma
Responsibilities
Own post-training pipelines for open-weight LLMs, design training systems using agent trajectories and user preferences, build and maintain datasets for AI agents, define measurable training rewards, run high-throughput rollouts, own distributed training workflows, establish model evaluation frameworks
Seniority
Senior, hands-on IC