Operational Architect, Business Applications

About this job posting\n

\n

We’re experimenting with hiring. Mainstream candidates won’t fit this role, so this is not a mainstream process.

\n

\n

If this role fits, apply through LinkedIn.

\n

About the role\n

We are not hiring a traditional software engineer.

\n

\n

We are hiring an Architectural Operator—someone who designs systems that produce software, not someone who just writes code.

\n

\n

With AI (“vibe coding”), code generation is no longer the constraint. The bottleneck is making what gets produced semantically and syntactically correct, reproducible, and architecturally sound.

\n

\n

Your job is to define systems that rapidly generate high-quality software.

\n

You will build and evolve agent-driven systems… that build and evolve agent-driven systems… that execute real business functions.

\n

\n

This seat is scoped to business applications. The agents and manufacturing systems you ship must produce working applications that run the company and our clients—origination-style workflows, data ingestion and reconciliation, operational back-office, QC, CRM/task systems, and the glue that replaces SI-style “pulling teeth” work. If the system doesn’t change how a business function actually runs, it doesn’t count.

The pre-interview process (mandatory)\n

\n

To qualify for an interview, you must build and submit an agent system. This is not theoretical.

\n

After you apply on LinkedIn, we will send a set of use cases. You pick one.

\n

\n

Example use case (illustrative):

\n

• \n
• Monitor Gmail and identify work-related emails
• \n
• Convert emails into structured tasks
• \n
• Use Asana as the system of record
• \n
• Suggest or generate responses
• \n
• Maintain state across systems and workflows
• \n

\n

\n

For this role, prefer a use case that produces a business application outcome (a runnable workflow that replaces or runs a real operational process)—not a content/marketing demo.

What we evaluate (all four at once)\n

\n

1. Speed (vibe coding in practice)

\n

\n

Build with AI-assisted development (Cursor or similar). Record a video of the full build (we expect ~1–2 hours, not a three-day film). We watch how fast you move, how you work with AI, and how you structure prompts and iterations. Velocity with control—not polish theater.

\n

\n

2. Architecture first

\n

\n

Define structure before you generate. Consistent patterns, not ad hoc code. Event-driven flows, clear orchestration (Step Functions or equivalent), defined data movement. Submit a short architecture overview (diagram or written) in the form of the agent’s skill(s).

\n

\n

3. Runnable, deployed system

\n

\n

Clean, understandable, runnable without friction, deployed (cloud or equivalent). We should run it end-to-end without manual fixes.

\n

\n

4. Reproducibility + iteration

\n

\n

On video, make a meaningful change (different account, different data structure, extended behavior) and regenerate. Example: monitor a different inbox; send tasks to a different Asana workspace. We then try our own changes to see if the system holds.

\n

Evaluation and outcome\n

\n

Automated review plus human review. We score architectural quality, speed and execution style, reproducibility, and whether the thing is actually useful as a business application.

\n

If selected: interview.

\n

\n

If the system is exceptional and usable: offer, plus a signing bonus for the work itself. If you build something valuable, you get paid for it.

What you’ll do on the team\n

\n

1. \n
2. Design software manufacturing systems — guardrails, enforced patterns, reproducible output at scale.
3. \n
4. Architect data-first systems — integrity and flow; graph databases, event-driven architectures, distributed pipelines; resilient and observable.
5. \n
6. Build and evolve agents — real business functions; encode logic, workflows, and decisions; raise output quality.
7. \n
8. Replace work with systems — find manual workflows, replace them with agents, cut operational overhead.
9. \n
10. Operate at high velocity — AI as a core tool; fast without becoming fragile.
11. \n

\n

\n

Business-application outcomes we care about: agents that ingest and reconcile operational data, stand up internal tools, wire legacy systems together, and ship functions a company can run next week—not slideware.

Technical environment\n

\n

AWS (primary): Lambda, Step Functions, EventBridge, SQS, SNS, DynamoDB

\n

\n

Serverless-first: stateless, event-driven; no long-running services unless necessary

\n

\n

Data: graph databases (Neo4j or similar), batch + streaming pipelines, strong observability

Principles\n

\n

• \n
• Code is disposable. Architecture is permanent.
• \n
• Reproducibility > reusability
• \n
• Speed without structure creates failure
• \n
• If it can be done twice, systematize it
• \n
• The best system is one you never need to fix
  
• \n

Who fits\n

\n

You think in systems, not tickets. You design patterns instead of rewriting the same code. You move fast and treat AI as a default tool. You care about correctness. You want to eliminate work, not create more of it.

Compensation philosophy\n

\n

We pay for fast, correct execution, systems that don’t need rework, and output that removes future effort. Highest-value work is done once and not revisited.

Why this role exists\n

\n

Business functions should be performed by agents. Agents should be generated by agents, not handwritten. Architecture—not code—is the asset.

\n

\n

If you can build systems that build business applications, apply on LinkedIn.