# Lead Product Software Architect — AI & Data
## Why this role exists
We're building AI-powered capabilities directly into customer-delivered software — not demos, not labs. This role owns the architecture that turns our data estate into an AI-ready product platform and makes AI features reliable, governable, secure, and scalable in production. You'll lead the modernization of our product data estate (schemas, pipelines, contracts, governance, and access patterns) so we can ship AI/ML and GenAI capabilities quickly and safely.
## What you'll own (outcomes)
- A clear AI + Data reference architecture that product teams can execute without heroics: from ingestion → curation → feature/embedding layers → serving → monitoring.
- A modernized data estate that supports rapid iteration: schema evolution, lineage, quality gates, and scalable access patterns (batch + real-time/event-driven where needed).
- AI capabilities that are production-grade: measurable quality, observable, performant, fully automated deployments, governance, and cost-optimized.
## What you'll do (Responsibilities)
### 1) Architect AI-enabled product capabilities (customer-facing)
- Translate business goals and product requirements into end-to-end architecture for AI features (e.g. predictive ML, recommendations, GenAI, agentic workflows).
- Define integration patterns between product services, data systems, and AI components (APIs, including MCP/A2A, ARG, events, model/agent serving, evaluation harnesses).
- Evaluate NFR tradeoffs and ensure delivery adherence (e.g. latency, cost, security, resiliency, and maintainability).
### 2) Modernize the data estate to be AI-ready
- Lead modernization of legacy data estates into a governed, scalable architecture (lakehouse/data mesh patterns, curated layers, data products, and contracts).
- Drive improvements in data quality, lineage, metadata, and discoverability — treat data pipelines as software (versioning, testing, CI/CD).
- Establish canonical models/semantic patterns that support analytics and AI/ML workloads (features/embeddings, training/serving parity).
### 3) Operationalize AI (MLOps/LLMOps) the "paved road" way
- Define standards and reusable patterns for: feature stores, model registries, experiment tracking, promotion workflows, drift monitoring, and retraining.
- Build reference implementations and enable teams to ship features repeatedly — moving from PoC to governed production delivery.
- Own architectural testing/validation practices