What you'll own
Technical direction for a swarm building a product every Pulse cycle. System design, integration strategy, data architecture, infrastructure decisions, and increasingly, how Amplify itself gets applied to the build. You treat architecture as a product asset: what you build today either compounds or forecloses what ships three cycles from now. You make the calls that keep a product coherent across cycles instead of a stack of one-off fixes.
How we build
Critical Propulsion pairs senior consultants with AI agents inside Amplify, our agentic delivery framework, to build products in 5-day Pulse cycles. Consultants own the judgment: what to build, what ships, which gate a release has to clear. Agents handle the volume: code generation, testing, the repeatable work a team of ten used to grind through.
Some engagements close out at delivery. Others keep going, the swarm shipping against a live roadmap cycle after cycle. Either way, week 1 is working product, not onboarding.
What we expect
- You weigh technical debt against roadmap speed as a product decision, not a hygiene issue. You can tell the difference between debt worth taking on this cycle and debt that quietly taxes every cycle after it.
- You make architecture decisions that hold up under 5-day delivery pressure, and under a product's second, third, and tenth cycle. No design that impresses in a review and buckles the first time requirements shift.
- You work cross-functionally, with real capability across data engineering, cloud infrastructure, and application architecture. You know how a decision in one domain ripples through the others.
- You embrace agentic development. You delegate real work to AI agents inside Amplify, review their output critically, and iterate fast. This is how the swarm builds now.
- You know which architectural decisions to delegate to agents and which ones need your judgment.
- You partner with the Product Delivery Lead before scope gets committed, not after. If something in the roadmap isn't buildable in a Pulse cycle, that's a conversation in the banding meeting, not a surprise in week three.
- You talk to the client directly when the technical direction is in question. You bring a better option, not a committee.
- You're a contributing architect. You set direction and you build. You write production code alongside the swarm, not just diagrams and decision docs.
- You operate autonomously. No status meetings that could be a message. No decks that could be a decision.
What we don't filter on
- Years of experience as a number. If you can set technical direction for a swarm shipping a product in 5-day Pulse cycles, the number is irrelevant.
Ideal candidates bring…
- Deep experience with Azure cloud services, including infrastructure as code, networking, and platform services in production.
- Experience designing data platforms, pipelines, or analytics infrastructure at scale, using tools like Azure Data Factory, Synapse, Databricks, dbt, or equivalent.
- Fluency in Python, Node.js, or TypeScript at a systems level, not just application code.
- Hands-on experience with agentic AI tooling: Claude Code, Cursor, Copilot Workspace, or similar. Better still if you've shaped how a team uses these tools, not just used them yourself.
- Experience designing systems where AI is a first-class component, not a bolt-on.
- Comfort with spec-driven design. You've written specs a swarm can build from in a single Pulse cycle without ambiguity or rework.
- Background making architecture decisions with a client in the room, not just an internal stakeholder.
- Prior experience on small, autonomous teams. You've set direction for a 3-4 person crew, not managed a 15-person org chart.
What you get
- A team where everyone builds. No layers of management between you and the work.
- AI agents as real teammates inside Amplify. You'll shape how the swarm uses agentic tooling, not just consume it.
- Pulse cycles that create rhythm without sprint overhead (KT +75%, Rework +60%, Mgmt +45% in typical sprint models).
- Product impact you can see. Your architecture decisions show up in production within days, and hold up across the cycles that follow.
- Competitive comp
How to apply
Send us something that shows how you think about systems. A design doc, an architecture decision record, a write-up of a hard technical call you made. Skip the cover letter unless you actually want to write one.
Let's have a real conversation. No fluff.
