AI-DNA Site Reliability Tech Lead

By the time you reach the alarm, the AI agents you architected have already validated hypotheses against years of prior incidents, parsed the logs and code paths, found the failure pattern, and proposed — or applied — a remediation inside policy guardrails. You own the reliability of one product family of an AI-native community and social engagement platform that the world's largest brands run their customer relationships on — where one outage is a brand-damaging event for a Fortune 100 company, so the standard is set there. You architect the agents that operate it, hold the quality bar that keeps AI slop out of production, and take the bridge as incident commander when it is down. Most of your week is in the agent code and the incident — and you are the most battle-tested engineer on your surface.

What you will be doing

• Own platform reliability for one product family — uptime up; MTTR and time-to-ack on Sev1 down week over week; SLAs held. You are the technical owner and incident commander; the customer relationship is carried at the executive level.
• Architect the AI agent system that operates your platform — pre-triage, change-validation gates, permanent-fix chase, RCA drafting, auto-healing — and decompose it into work a small senior team can build and maintain. The harness is the product.
• Own the safe path to production — every change, not just every incident: deploys, cost-optimization runbooks, and config changes from other teams ship through your quality gates and tested rollbacks.
• Hold the quality bar — the approval gate on complex changes and the internal QC on every RCA. Nothing reaches production, or leaves your team, that you would not put your own name on.
  
  

What you will NOT be doing

• Becoming a people manager. You own the team's technical output and quality bar — not their comp, careers, or hiring. If your next move is HR-style management, this is not it.
• Owning the customer relationship. You make the platform unbreakable and the RCA correct; carrying it to the customer is owned at the executive level.
• Growing the team to do what AI should. Add headcount to cover agent gaps and you are off-strategy — fix the AI, not the org chart.
• Treating this as a credential or a résumé line. The bar is ownership, depth, and obsession — not titles. If you have never been deeply, sustainedly obsessed with a hard problem, this is a bad fit.
  
  

Responsibilities

• Deliver the reliability outcomes — own them like a founder. Uptime up; MTTR and time-to-ack on Sev1 down week over week; SLAs held. When it is down, that is your problem; you do not sleep peacefully through a customer outage.
• Command the incident. The senior technical voice in the room — diagnose, decide, restore, and hand a precise picture up the line. The team escalates to you; you escalate when the blast radius warrants.
• Architect the agent surface — and decompose it for the team. New agents that catch a class of incident before it recurs, existing ones tuned higher, guardrails added when production teaches you something the spec missed. Most of your week is here.
• Own the RCA discipline — root cause, not symptom, prevention that ships. Symptom separated from cause; you QC every RCA and write the hardest. You watch two clocks — time-to-RCA and time-to-close-the-actions. An RCA whose fixes never ship does not count.
• Guard production on every change. Deploys and config changes from across the company ship only through your gates, with tested rollbacks; you hold the line when a change is not safe yet.
• Multiply the team, and write so the work survives you. Fix the underlying agent when it fails and share the fix as reusable rules, context, and skills; document so an agent can retrieve it and the next person is not starting cold. If it works only when you are online, you have not finished it.
  
  

Requirements

• 7+ years operating SaaS at scale, 2+ as a Tech Lead / Staff / Principal leading senior engineers — a senior-IC + technical-leadership track, not people management.
• Extreme ownership, fully self-directed. You run your platform like your own business, and you can answer "the hardest you have ever worked on something" with a story we can probe for an hour. Nobody checks your work; you challenge a standard with a better idea but never quietly ignore it. If you need babysitting, this is the wrong team.
• The deepest production scars on the team. The most battle-tested engineer on your platform — you have lived the worst outages and know the failure modes cold. Deep AWS at scale (multi-AZ, multi-account; not Azure-only, not greenfield), you run production change with gates and tested rollbacks, and you investigate to the actual root cause. You could run it all by hand — the point is you bake it into the agents.
• AI-native operator who multiplies. You delegate whole units of work to agents that verify their own output, fix the underlying capability when they fail, and share it so the team levels up — measured on leverage, not AI-usage %. Advanced across agentic tools (Claude Code, Codex, Warp, custom agents); when a new model ships, you are testing it that day, not next quarter.
• Fluent / advanced English — precise under pressure, in writing and on the bridge.
• US-morning overlap (≈13:00–17:00 UTC), plus on-call rotations and some shift coverage — reliability is a 24/7 outcome and you carry your share.
• OFAC-clear country of residence.
  
  

Nice to Have

• AWS Solutions Architect – Professional — or a production track record that makes the certification redundant. Scars over badge.
• Pioneer voice in agentic SRE / AIOps — original posts, talks, or open-source work. Followers and reposters are not what we are looking for.
• A documented obsession with one hard thing outside of mainstream work — depth matters more than the title on the latest role.
• Modern observability / incident-response stacks (Grafana, Prometheus, OpsGenie, PagerDuty, Datadog) and Azure exposure alongside your AWS depth.
  
  

What you will learn

• You will be one of the engineers defining what agentic SRE looks like in production — for a platform the world's largest brands depend on, running a team where AI agents are full members and the surface you build is its force multiplier. You will leave with the kind of agent-ops experience the rest of the industry is still trying to hire.
  
  

Working conditions

• Enterprise scale, startup cadence. Large-enterprise customers, footprint, and contractual stakes; a startup operating model — outcomes measured week over week, decisions in days, the playbook rewritten as the field moves. If you need a year to land the framework from your last role, this is not it.
• Remote, async-first, global — top 1%, AI-First by default. Hire from anywhere with US-morning overlap; no office; small and senior, no L1 / L2 beneath you. AI is how this team thinks and operates, not a tool it is evaluating.
• No token or tooling limits. The harness is the product, and we resource it that way — if the right answer is more compute, a better model, or a tool we have not bought, we buy it.