Data Scientist (LLM)

What You'll Do:

• Train and fine-tune LLMs using multi-GPU and distributed setups.
• Analyze model performance, debug failures, and implement improvements.
• Work with curated and synthetic datasets, including classification and generation tasks.
• Design experiments, track results, and iterate quickly with tools like MLflow.
• Write clean, production-ready code and collaborate via GitHub.
• Push boundaries: think about architecture, memory efficiency, scale, and cost.

What We’re Looking For:

• LLM understanding: You know how these models work under the hood - transformer internals, tokenization, embeddings, etc.
• Stats and ML fundamentals: You have a solid foundation in statistics, machine learning, and optimization.
• Coding skills: You write Python well. You’ve worked with PyTorch or JAX. You don’t fear Bash or git.
• Training experience: You’ve trained models beyond notebooks. Bonus if you’ve worked with mixed precision, DeepSpeed, or multi-node training.
• Systems mindset: You understand trade-offs - throughput vs memory, latency vs accuracy.
• Good judgment: You know when to read a paper, when to read a stack trace, and when to rewrite the dataloader or part of the LLM architecture.

Bonus Points:

• Experience with Hugging Face Transformers, Datasets, Accelerate.
• Familiarity with Kubernetes, Ray, or custom training infra.
• Exposure to embeddings, classification tasks, or token-level losses.
• Ability to mentor or guide junior researchers or engineers.

How We Hire:

• Online assessment: Online assessment: technical logic and fundamentals (Math/Calculus, Statistics, Probability, Machine Learning/Deep Learning, Code).
• Technical interview: dive into theory and reasoning (no code).
• Cultural interview.


If you are not willing to take an online quiz, do not apply.