Frontend Engineer, Vision

🏢 About Sarvam

Sarvam is building India's full-stack sovereign AI platform — covering research, model development, infrastructure, and production applications. The company's focus is on making AI genuinely work for India, including its languages, documents, and institutions. Backed by Lightspeed, Peak XV, and Khosla Ventures, Sarvam works with enterprise and public-sector partners including Tata Capital, SBI Life, CRED, IDFC, and LIC.

🎯 The Role

The Vision team builds and runs in-house vision-language models for OCR and structured extraction from Indian documents at population scale — PAN, Aadhaar, bank statements, GST filings, insurance reports, contracts, legal filings, and more. Document intelligence is only trustworthy when a human can see what the model saw. As the team's frontend owner, you will build the interfaces that make that possible: document viewers, correction workflows, schema builders, evaluation dashboards, and the developer console enterprise customers use to run pipelines at scale.

🛠️ Tech Stack

• React + TypeScript
• Canvas and SVG for custom rendering
• PDF.js (or equivalent) for in-browser PDF rendering
• WebSockets / SSE for streaming
• Go + Python backend (Temporal, Kubernetes, PostgreSQL, Redis)

💼 What You'll Do

• Build the document review experience: PDF and image rendering, page navigation, bounding-box overlays, confidence highlighting, and side-by-side source-to-extraction linking
• Build human-in-the-loop correction workflows — keyboard-driven review, field-level editing, approve/reject queues — and wire corrections back into the evaluation loop
• Build an extraction schema designer where users define, test, and version the structured output they want from a document type
• Build internal evaluation and observability dashboards tracking accuracy by field and document type, latency, cost, and model comparisons
• Handle real-time and long-running state well: streaming partial results, job progress, optimistic updates, resumable sessions
• Own frontend performance on genuinely heavy documents — virtualisation, canvas rendering, lazy loading, memory discipline
• Shape API contracts with backend engineers; you define them, not just consume them
• Set the frontend standard for the team: component architecture, typing, testing, accessibility

✅ What You'll Need

• 4–5 years building production frontend applications with strong end-to-end ownership
• Deep React expertise — hooks, rendering behaviour, state management at scale, performance profiling
• Strong TypeScript; you use the type system to make bad states unrepresentable
• Real command of the browser platform: layout, events, memory, rendering; comfortable with Canvas or SVG for custom rendering
• Experience with complex, data-dense product surfaces — dashboards, editors, annotation or review tools — not marketing pages
• Solid async state, caching, polling, streaming, error and retry handling instincts
• Design sensibility to ship clean UI from a rough direction without pixel-perfect mocks

🌟 Nice to Have

• Experience with PDF rendering in the browser (PDF.js or equivalent) or annotation/labelling tools
• Familiarity with WebSockets, SSE, or other streaming transports
• Exposure to AI/ML product surfaces — human-in-the-loop review, model output inspection, eval tooling
• Ability to drop into a Python or Go backend when it unblocks you
• Design systems ownership or strong accessibility practice