Director/Principal Engineer, Data & AI Platform

Arlo•Milpitas, CA

13h•$225,000 - $300,000

About The Position

Arlo is looking for a senior technical leader to turn our data platform into an AI-native internal product — one with real consumers, a clear owner, and a roadmap. The difference this role makes should be measurable: Any team at Arlo can find a trusted metric or dataset without asking around. New data products get built on shared primitives instead of from scratch. Dashboards get retired because a governed, self-service version replaced them. AI tools and agents can query business data safely, because access and metadata were designed for that from the start. You will close the gap between “we have data” and “we can act on data” — through the architecture, primitives, and defaults you build. This is a software engineering and architecture role first. You’ll write and ship production code, own the tests and CI around it, and operate what you build — designing the interfaces, libraries, and primitives other teams build on, with the versioning, backward-compatibility, and reliability guarantees that implies. Your first six months should produce running systems, not documents. If your recent work has been mostly diagrams, roadmaps, and specs, this isn’t the right fit. You’ll work horizontally across Data Engineering, Analytics, Cloud Infrastructure, and Enterprise Applications as the technical reference point for how teams model, catalog, and expose data, and you’ll regularly lead cross-team technical initiatives.

Requirements

AI and LLM systems. Production experience building with LLMs — retrieval and natural-language interfaces over structured data, context and prompt design, and the evaluation, guardrail, and cost controls that make them safe to put in front of non-technical users. Shipped and operated, not prototyped.
ML as a platform capability. Stood up from the ground up and made usable by people who aren’t ML engineers — feature definition, training and serving paths, evaluation, and drift and quality monitoring.
Lakehouse depth in production. Unity Catalog, Delta Lake, medallion architecture, Lakehouse design patterns — with metadata, lineage, and governance designed in rather than bolted on. We run Databricks; deep production experience on a comparable platform transfers.
Heterogeneous data integration. Production experience reconciling data across stores with different consistency models, key spaces, and access controls — identity resolution and referential integrity without shared keys, and query patterns that protect operational systems from analytical and agent workloads.
Enterprise data architecture. Dimensional and semantic modeling, master data and ownership models, data contracts, and taxonomy design across multiple business domains and stacks. Not a single-team data mart — something several consuming teams depend on.
AWS at scale. AWS data services (DynamoDB, Redshift, Glue, or equivalents) in production.
Programmatic data access. Access control and audit design for non-human consumers, and familiarity with MCP or comparable tool-callable interfaces.
Depth and range. 10+ years in data engineering or data platform roles, including production ownership of systems multiple teams depend on.
Coding. Strong SQL and Python (or another general-purpose language), plus the testing, code review, and production ownership that go with shipping code other teams depend on.
Adoption without authority. A track record of getting a platform capability adopted by teams that don’t report to you.

Nice To Haves

Data catalog and governance tooling (Collibra, Atlan, or similar).
A regulated or compliance-sensitive data environment.
Replacing dashboard and reporting sprawl with governed self-service.
Oracle EBS, Amplitude, Klaviyo, or similar enterprise and product-analytics sources are a strong plus.

Responsibilities

Treat the platform as a product. Know your consumers — analysts, engineers, applications, AI tools — and where today’s experience falls short. Drive a roadmap measured on adoption, trust, and time-to-answer, and make the case for foundational investment with evidence of consumer need.
Ship the shared primitives other teams build on top of — not one-off deliverables, and not starter kits that fork and diverge.
ML and AI as a capability. Ship the first governed features analysts can apply to trusted data products themselves: anomaly detection over metrics, forecasting, segmentation, natural-language query against the semantic layer. Build them as repeatable capabilities with evaluation, monitoring, drift and cost visibility from day one — not a portfolio of bespoke models.
Design how AI tools and agents reach the platform (MCP or comparable) as a core interface held to the same query-safety and permission guarantees as any other consumer — not a side channel around the controls.
Solve the cross-store problem. Arlo’s data lives in systems that were never designed to be joined — DynamoDB for operational and device data, Databricks for analytics, plus Oracle EBS, Amplitude, and Klaviyo. Design the identity resolution, referential integrity, and consistency semantics that let a consumer trust a join across them.
Define query-safe access patterns so a dashboard, application, or agent can’t overwhelm an operational store or quietly return a wrong answer — and a permission model that holds across systems with entirely different native access controls.
Architecture. On that foundation, design Arlo’s semantic and metric layer and the catalog, so business terms, metric definitions, and ownership are consistent, discoverable, and reusable.
Define data product contracts — schemas, ownership, freshness and quality SLAs, access patterns — including what’s safe to expose to dashboards, applications, and AI tools.
Governance. Build metadata, lineage, and access-control models into the platform itself, so governance is a property of the system rather than a manual review.
Define what “trusted” means for a data product at Arlo, and make that trust visible and checkable.
Separate foundational investment from one-off requests, and partner with Data Engineering, Analytics, and AI engineering leadership on build-vs-buy and sequencing.