AI/NLP Engineer

Sarvam AI is on a mission to build AI for India's 1.4 billion people across 22 official languages. We are hiring an exceptional AI/NLP Engineer to work on our foundational language models and speech recognition systems.

This is a rare opportunity to work at the frontier of AI research applied to underserved languages. You will pre-train, fine-tune, and evaluate large language models on Indian language data, contributing to one of India's most ambitious AI projects.

• Pre-train and fine-tune transformer-based models for Indian languages
• Build data pipelines for collecting and cleaning multilingual datasets
• Evaluate model performance using linguistic benchmarks

• Design and implement NLP model architectures for Indian languages
• Build data collection, cleaning, and preprocessing pipelines
• Conduct model training experiments and analyze results
• Optimize model inference speed and memory usage
• Collaborate with linguists and annotators on data quality

• 3-6 years of NLP/ML engineering experience
• Strong Python skills and experience with PyTorch or JAX
• Experience with transformer architectures (BERT, GPT, T5, etc.)
• Knowledge of speech processing (ASR, TTS) is a big plus
• Experience with Hugging Face Transformers library
• Strong mathematical foundations in linear algebra and probability
• Publications in NLP or speech conferences are a plus