IgniteTech
Website:
eti.com
Company:
https://www.linkedin.com/company/ignite-tech
Seniority: Not Applicable
Industries: Software Development
Job details:
When the alarm fires, you are the first responder — but by the time you reach it, the AI agents you built and maintain 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. Your job is the shift and the agents that run it: take the page, resolve what the agents could not, and make sure the next page of that kind handles itself. You operate 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 are a senior SRE who could run the whole operation by hand; the way this team works, you bake that expertise into the agents and runbooks instead of spending your life in a ticket queue.
What you will be doing
- Own the shift as first responder — take the page, triage and resolve customer-impacting incidents on the platform you cover, and escalate to the tech lead when the blast radius warrants.
- Execute production changes safely — deploys, cost-optimization runbooks, and config changes from other teams run through the quality gates with a rollback ready before you start, and you pull it the moment a change goes off-plan.
- Build, operate, and maintain the AI agents that do the operations work — pre-triage, change-validation gates, permanent-fix chase, RCA drafting, auto-healing — and write the AI and human runbooks they run on. The harness is the product.
- Execute the ticket work — but measured on how well your agents make it scale, not how many tickets you personally close. Every one-off you solve by hand, you generalize so an agent handles the next one.
What you will NOT be doing
- Closing tickets as the goal. Throughput is not your scoreboard; the leverage your agents create is. Hours into a queue instead of into the agents that drain it means you are off-strategy.
- Waiting to be told what to do. You pick up the page, find the next gap, and ship the fix. Nobody assigns you your day.
- Solving the same problem twice by hand. A one-off you do not generalize is unfinished work.
- 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
- Own the shift and the outcome on it — like a founder. You are first on the page. When something is down on your watch, that is your problem; you do not sleep peacefully through a customer outage.
- Build, operate, and maintain the agents that run the operation. 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. You build within the architecture the tech lead sets — and push back with a better idea when you have one.
- Protect production every time you touch it. When you run a deploy, a cost-optimization runbook, or a config change, you go through the quality gates, keep a tested rollback ready, and pull it the moment the change goes off-plan. You are a guardian of uptime, not only a responder to its loss.
- Find the root cause — not the symptom — and close the loop. A disciplined, structured investigation that separates symptom from cause; then the part most teams skip — identify the prevention, build it, and track it to done. The tech lead reviews your RCAs; you write them to a bar that needs little correction. An RCA whose fixes never ship is theater.
- Generalize relentlessly. You may solve a one-off to restore service, but you do not stop there — you turn it into an agent, a runbook, or a guardrail so the next occurrence handles itself. The scoreboard is what the agent surface can carry, not how fast you close tickets.
- 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; encode every procedure so an agent can retrieve it and the next responder is never starting cold. In an async team, an undocumented fix is a fix that doesn't exist.
Requirements
- 5+ years operating SaaS at scale as a senior, on-call SRE — real first-responder time on production incidents, not just project or build work.
- Extreme ownership, fully self-directed. You run your shift 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 assigns your day or checks your work; you challenge a standard with a better idea but never quietly ignore it. This team does not babysit — if you need to be managed, it is the wrong fit.
- Deep production scars — outages, incident management, hands-on troubleshooting. First on the bridge, diagnosis under pressure, the failure modes known from having lived them. Deep AWS at scale (multi-AZ, multi-account; not Azure-only, not greenfield); you run production change with gates and tested rollbacks and investigate to the actual root cause, symptom separated from cause. You could run it 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.
- AWS Solutions Architect – Associate (or higher) — or a production track record that makes the certification redundant. Scars over badge, but a useful floor at this level.
- Fluent / advanced English — precise under pressure, on the bridge and in writing.
- Shift coverage is core, not occasional — on-call rotations and the shift window you are hired into, with the time-zone overlap that window requires. Reliability is a 24/7 outcome for the whole team, and you carry the share in your geographical area.
- OFAC-clear country of residence.
Nice to Have
- Pioneer or builder voice in agentic SRE / AIOps — original posts, talks, open-source, or agents and tools you have built and shipped. - - 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.
- Background in multi-tenant B2B SaaS at scale — community, social, customer-experience, observability, or AIOps platforms.
What you will learn
- You will become one of the engineers who actually built agentic SRE — not read about it — on a platform the world's largest brands depend on, where AI agents are full members of your team and the runbooks you write run themselves. You will leave with hands-on agent-ops and incident 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.
- 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.
Click on Apply to know more.
