Senior Engineer (.NET + Python) with AI-augmented

You’ll build the intelligence and autonomy layers on top of a solid architecture: .NET 9 modular monolith + Python/FastAPI services on GCP (Pub/Sub, Cloud Tasks, BigQuery, Supabase/PostgreSQL) with live Meta Ads API and Shopify GraphQL integrations.
In the first 90 days you’ll deliver three defined sprints: Shopify attribution + benchmarks (Sprint 1), Creative Intelligence with embeddings + LLM fallback (Sprint 2), and an agentic loop with an Action API that executes approved Meta Ads actions (budget scaling, pause, deploy) with audit/audit trail (Sprint 3).

Product & Environment

We’re building a system with a modern stack and live integrations, where your decisions affect real ad accounts:

• C#/.NET 9 modular monolith, Python/FastAPI microservices, Next.js 16, Supabase/PostgreSQL, BigQuery, Google Cloud (Pub/Sub, Cloud Tasks, GCS)
• Integrations with Meta Ads API and Shopify GraphQL
• 20 repositories, ~95k LOC, event-driven pipeline

How We Work

Claude Code is not optional. We use AI as a force multiplier: we measure output, not hours. The code must work, tests must pass, and APIs must behave correctly.
The wrong candidate either refuses to use AI tools (“I prefer to write it myself”) or over-trusts them (“Claude Code wrote it so it must be correct”).

Responsibilities

• Build features in .NET (core API) and Python/FastAPI (pipeline / categorization / agent services)
• Integrate with Meta/Shopify and work on real budget/campaign changes
• Work with PostgreSQL: schemas, migrations, indexes, query optimization
• Strong testing culture: unit/integration/E2E where needed

Requirements (Non-negotiable)

• Production C# / .NET 8–9: EF Core migrations, DI patterns, middleware; able to read a modular monolith without “hand-holding”
• Experience with GraphQL resolvers (HotChocolate is a plus, but any GraphQL server is fine)
• Production Python / FastAPI: async/await, Pub/Sub message handling, gRPC; integrations with AI APIs (OpenAI/Anthropic, etc.)
• API integrations: integrated with at least two complex third-party APIs (Meta Ads / Google Ads / Shopify GraphQL / Stripe, etc.), including rate limiting, pagination, webhooks, versioning
• AI-augmented development: daily use of Claude Code / Cursor / Copilot (or similar) + understanding where they fail
• Testing discipline: tests are part of the process, not “later”

Strong Plus

• Ecommerce/DTC (Shopify, attribution, ROAS, NC vs returning customers)
• Experience with embeddings / cosine similarity / clustering / recommendations
• Event-driven architecture (Pub/Sub or Kafka), idempotency, DLQ, eventual consistency
• Hands-on GCP (Cloud Run, Cloud Tasks, Pub/Sub, BigQuery, GCS)
• Experience preparing a SaaS/agency for exit (focus on quality, documentation, system autonomy)

What You’ll Build (First 90 Days)

We have 3 detailed sprints — no ambiguity or fluff — with clear deliverables from Day 1.

Sprint 1 (Weeks 1–2): NC Attribution + Shopify Integration

• Customer/Order entities; NC classification (customer.numberOfOrders == 1)
• Channel attribution via Shopify customerJourneySummary + click ID parsing (fbclid/gclid/ttclid)
• EWMA benchmark fixes + unit tests; CM1/CM2/CM3 contribution margin; Benchmarks page API

Sprint 2 (Weeks 3–5): Creative Intelligence + Testing Arena

• Attribution Confidence scoring (High/Med/Low/Dark)
• Efficiency Score (0–100), zone classification (Winner/Functional/Testing/Kill Zone)
• Persona/Angle categorization via embedding similarity + LLM fallback
• Creative Matrix data layer, Scale Opportunities endpoint

Sprint 3 (Weeks 5–8): Agentic Loop + Owner Independence

• Decision Engine (rules-based action queue + priority scoring)
• Action API: execute approved actions via Meta Ads API (budget scaling, pause, deploy)
• Decision logging + audit trail
• Integrations: Nano Banana creative generation, MILA AI assistant, brand onboarding

Not a Fit If

• You think AI coding tools are a fad or “beneath senior engineers.” Our workflow is built around them.
• You need architecture hand-holding. Sprint scopes are detailed, but you must make real-time decisions on data modeling, API design, and integration patterns without waiting for approval.
• You only work in one language. This role requires fluency in both C# and Python. TypeScript is a bonus, not required for this scope.
• You’ve never integrated with complex third-party APIs. If your experience is mostly internal CRUD apps, Shopify/Meta edge cases will slow you to a crawl.
• You ship code without tests. We have 440 lines of untested financial math (P0 fix). Testing is core to the job, not “if there’s time.”
• You need a detailed spec for every function. We define WHAT to build and WHY — you decide HOW.
• You want a strict 9–5. This is a high-intensity engagement with a clear deadline. We’re not asking for 80-hour weeks, but we need someone who ships consistently and communicates proactively when blocked.