Teams at Sarvam

Own product roadmaps for ASR, TTS, and model evaluation. Coordinate data collection, training cycles, and releases. Build evaluation frameworks and run experiments (offline, A/B, human) to improve accuracy, latency, MOS, and robustness. Ensure scalable deployment, monitoring, and translate business needs into measurable model goals.

Deploy alongside clients to transform heterogeneous multimodal data into semantic structures for AI reasoning. Design ontologies, segmentation, embedding and retrieval strategies, collaborate on ingestion pipelines, evaluate retrieval and reasoning performance, and translate insights into product and engineering signals.

Design and maintain domain-specific evaluation frameworks for LLMs and AI systems: define quality metrics, run pre/post-deployment evaluation cycles, build dashboards, find failure modes and distribution shifts, operationalize automated eval pipelines with MLOps, manage datasets and annotation workflows, and publish quality reports.