Principal AI Architect at Reliance Infosystems

The mission of this role

• You are the Studio's technical authority. Every product it ships carries your architectural signature: governed, evaluated, secure, and economical to run. You will teach the academy's hardest weeks yourself, review every venture from its solution gate onward, and set standards that make fourteen early-career builders produce software a CIO will trust. This role demands current, demonstrable depth on the Microsoft AI stack agent platforms, data, identity and governance. Technical credibility is at the core of any studio, we will test for it to ensure candidates suitability.

What you will own

• Reference architecture and standards: multi-tenant, agentic, Azure-native product patterns — identity for agents, observability, evaluation-gated releases, kill switches — adopted by every venture.
• Quality gates: architecture and evaluation review of every venture from solution validation onward; nothing reaches a paying customer un-reviewed.
• The academy's technical spine: teach the foundations, agent-engineering and trust modules; run the architecture review bench; make teaching a first-class output, not a distraction.
• Unit economics: cost-per-run telemetry on every build from its first commit; consumption-aware design so pricing survives contact with reality.
• Responsible AI and security by design: with the fractional advisor, own the pre-screen and full review that keep products shippable in regulated sectors.
• Marketplace readiness: architect every product to qualify as a transactable cloud-marketplace offer, certification included.
• The hardest problems: stay hands-on — the gnarliest integration, the failing eval, the 2 a.m. pilot incident are yours to lead from the front.

Your first ninety days

• Reference architecture and engineering standards published and adopted; environments and CI/CD live for all squads.
• The academy's first technical modules delivered by you, with teach-back quality the cohort rates highly.
• One governed internal agent per squad in production — identity, observability and kill switch included — used daily by real staff.
• Evaluation harness live; the first release blocked by a failing eval (and the lesson taught from it).
• Cost-per-run telemetry flowing from every active build.

How success is measured

Performance is reviewed against the Studio's published scoreboard outcome-based, evidence-first, and visible to leadership monthly. Targets below are the twelve-month bar; trajectory reviews at months four, seven and ten.

Key performance indicator

12-month target

Measured via

Standards adoption

• 100% of ventures on reference architecture; zero un-reviewed releases
• Architecture review log

Production quality

• ≥5 customer pilots in production under full governance; zero critical security or AI-safety incidents
• Governance register · incident log

Evaluation discipline

• 100% of releases evaluation-gated; eval coverage visible per venture
• Release records

Unit economics

• Cost-per-run tracked on 100% of builds; unit economics signed before every paid pilot
• Venture gate packs

Cohort capability

• Certification ladder on schedule (~42 certifications year-one); every engineer shipping production code monthly
• Flight-hours ledger

Reusable assets

• ≥2 accelerators/patterns in the Studio library, reused across ventures
• Library audit

Marketplace readiness

• First product certified as a transactable marketplace offer
• Certification record

Teaching

• Core technical modules delivered; academy feedback ≥ agreed bar

Academy reviews

What you bring

• 5+ years of software engineering with principal/staff-level scope; you have been the final technical word on systems that mattered.
• You have shipped production AI systems LLM-powered products or agents with real users and can talk concretely about evals, guardrails, failure modes and cost management, not just frameworks.
• Deep, current Microsoft stack capability: Azure architecture plus the AI/agent platform family (agent runtimes, low-code agent tooling, the data/analytics platform, identity and governance for agents). Relevant certifications are welcome; live depth in an interview matters more.
• Multi-tenant SaaS architecture experience — security, isolation, metering and operability designed in, not bolted on.
• A genuine teaching record: mentorship, workshops or lecturing — and the temperament to enjoy it. A teaching audition is part of our process, and we mean it.
• Delivery into regulated industries (financial services, energy, health or similar) is a strong advantage.

It helps if you also bring: published works or a public technical voice; open-source contributions; experience taking a product through cloud-marketplace certification.