About the role
TL;DR
Own the architecture of Sarvam's vision model serving harness for national-scale document intelligence.
- •You will own the architecture of the serving harness for Sarvam's vision models, ensuring frontier-grade extraction quality at national scale within cost and latency budgets.
- •Key Responsibilities Own the end-to-end architecture of the OCR and extraction serving harness.
- •Design the accuracy harness for multi-pass extraction and ensembling.
- •Architect durable, resumable document workflows in Temporal.
- •Drive latency, throughput, and unit economics down deliberately.
- •Build the observability substrate for distributed tracing and metrics.
- •Requirements 5-6+ years in backend engineering, operating high throughput production systems.
- •Deep proficiency in Go and/or Python.
- •Strong distributed systems design experience.
- •Production experience with Temporal or an equivalent durable execution engine.
- •Kubernetes in production experience.
Required skills
GoPythonKubernetesPostgreSQLRedisREST APIMicroservices
Nice-to-have skills
Computer VisionLLMs
Domain expertise
ai
Tech stack
GoPythonKubernetesPostgreSQLRedis