Forward Deployed AI Engineer

Must Have

• We're looking for genuine production depth across data engineering and full-stack development — not surface familiarity with either.
• Data Engineering Foundation
• Data modeling and schema design — dimensional modeling, normalization trade-offs, and EDW/warehouse schema design you can defend.
• Hands-on data pipeline experience — ETL/ELT design across batch and incremental loads, built and maintained in production (not just SQL scripts on a schedule).
• Slowly Changing Dimensions (SCD) and change-data handling — knows the patterns and when each applies.
• dbt Experience— modular SQL transformations, tests, documentation, and incremental strategies.
• Advanced SQL and at least one modern data platform in depth (e.g., Snowflake, Databricks, or a comparable cloud warehouse/lakehouse).
• Data quality thinking — testing, validation, and lineage treated as first-class, not afterthoughts.
• Full-Stack AI Product Development
• Python as a primary language — services, automation, and data work alike.
• FastAPI — async REST API design, dependency injection, testing.
• A modern frontend, ideally Next.js — component architecture, SSR, state management, and real UX sensibility.
• PostgreSQL — schema design, query optimization, indexing.
• System design — can architect from a blank page: services, boundaries, trade-offs, and scale.
• AI-paired engineering — uses an agentic coding tool (Claude Code, Cursor, or comparable) as a genuine daily workflow accelerator, and can speak concretely to how.
• CI/CD and cloud deployment ownership on AWS or Azure, without heavy support.
• Ways of Working
• Comfortable in client-facing delivery — can represent TechTorch technically and translate between business and engineering.
• Customer-first mindset — anchors decisions in what the stakeholder is actually trying to accomplish, and can move fluidly between the engineer's view and the business owner's in the same conversation.
• End-to-end ownership instinct — takes a problem from discovery to production and owns the outcome, rather than passing it along at each handoff.