Senior Backend Engineer, Vision

About the role

We are looking for a Senior Backend Engineer to own the architecture of Sarvam’s vision‑language serving harness. You will build a production‑grade OCR and extraction platform that runs at national scale, balancing accuracy, latency and cost.

Key responsibilities

• Design and own the end‑to‑end architecture of the OCR and extraction serving harness, including API layer, orchestration, inference, post‑processing and delivery.
• Implement multi‑pass extraction, ensembling, confidence calibration and schema‑constrained decoding to achieve frontier‑grade quality.
• Build durable, resumable document workflows in Temporal with fan‑out, partial‑failure recovery and exactly‑once side effects.
• Manage inference serving: batching strategy, GPU pool autoscaling, multi‑model routing and cost‑per‑page optimisation.
• Develop observability: distributed tracing, per‑stage cost and quality metrics, SLOs, alerting and post‑incident analysis.

Required profile

• 5–6+ years of backend engineering experience, including on‑call responsibility for high‑throughput production systems.
• Strong expertise in Go and/or Python and solid judgement on language choice.
• Deep knowledge of distributed systems design: queues, workflow orchestration, idempotency, back‑pressure and graceful degradation.
• Hands‑on experience with Temporal (or equivalent) and Kubernetes in production, including GPU workload scheduling.
• Proven track record serving ML or LLM inference at scale, with batching, caching and latency budgeting.

Required skills

• Go
• Python
• Temporal
• REST
• Kubernetes
• PostgreSQL
• Redis
• Object storage
• OpenTelemetry