SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Sarvam AI is building India's full-stack sovereign AI platform, backed by Lightspeed, Peak XV, and Khosla Ventures. The company works with leading enterprises and public institutions including Tata Capital, SBI Life, CRED, IDFC, and LIC.
As Product Manager for Model APIs, you will own how developers and enterprises consume Sarvam's models across speech, language, and multimodal domains. You sit at the intersection of research and market, responsible for the complete API surface—specification, versioning, documentation, developer experience, and pricing—while also influencing the model roadmap itself.
Key responsibilities include: defining and owning the API product surface; shaping model roadmaps alongside research leads using usage data, evaluations, and customer signals; establishing quality standards (accuracy, latency, cost per call, robustness); owning evaluation as a product including benchmark suites and reporting; translating customer and developer feedback into concrete requirements; partnering with infrastructure teams on production rollout and reliability; driving end-to-end release planning with positioning and migration strategies; and communicating priorities across research, engineering, GTM, and leadership.
You should bring 3+ years building AI/ML products, developer platforms, or APIs where the underlying model was part of product decisions. You need working fluency with the model lifecycle (data, training, fine-tuning, evaluation, deployment, monitoring) without necessarily training models yourself. Strong analytical instincts, comfort with data-driven decision-making, and sound judgment on quality/latency/cost trade-offs are essential. You must be able to work closely with researchers and earn their trust.
Bonus experience includes speech (ASR, TTS) or LLM APIs with relevant metrics (WER, CER, MOS, TTFT, tokens per second); shipping and maintaining public or enterprise APIs; hands-on SQL, Python, or ML tooling for independent analysis; and exposure to Indian language AI, multilingual systems, or low-resource language evaluation.
