About the Role

This is a pure hands-on technical leadership role no people management. You own the technical direction of the AI platform: the systems that take models from experiment to production agentic run times, retrieval and context pipelines, inference serving, evaluation infrastructure, and the safety/cost guardrails around all of it. Changes reach production in minutes, and defects surface as hallucinated answers to customers, runaway inference spend, or leaked context across tenants.

You are the highest technical authority in the domain. You lead through architecture, code, and review not org charts. Your judgement must stay sharp enough to make final architecture calls and current enough to review any MR in the platform, in a field where "current" has a half-life of months.

What You'll Do:

- Review MRs , create a system for MR acceleration with a bar that catches nondeterminism handled as if deterministic, unbounded context growth, missing eval coverage on prompt changes, retry storms against rate-limited providers,



and PII flowing into prompts or logs and teach the team to catch them first.
- Design and defend one significant platform change (e.g., introducing a model-routing layer, migrating retrieval infrastructure, standing up an agent-execution sandbox, re-architecting credit metering) through design review, security review, and phased rollout behind flags.
- Materially improve evaluation infrastructure: regression suites that gate deploys, golden datasets, online quality monitoring — so "did the recent model/prompt make things worse" is answered by CI, not by customers.
- Ship code on the critical path yourself — the hard 20%: the concurrency bug in the agent loop, the streaming backpressure issue, the tokenizer-boundary edge case.
- Serve as the design authority in technical interview loops (system design, architecture rounds) — assessing candidates, not owning the hiring pipeline.

📌 Technical Architect (Udaipur)
🏢 Fusion Business Solutions
📍 Udaipur