Principal Engineer, Data Quality Platform

INSTITUTIONAL FABRIC FOR THE DIGITAL ASSET MARKET Founded in 2018, Talos provides institutional-grade trading technology for the global digital asset market, powering many of the major players in the crypto ecosystem. Our mission is clear: to advance the mass adoption of digital assets by seamlessly connecting institutions to the digital asset ecosystem. We are committed to building the most innovative and trusted platform in the world, supporting the entire trading lifecycle. At Talos, you’ll find an environment that champions kindness and respect, values diverse perspectives, and upholds inclusivity at every turn. We believe every member of our team brings invaluable insights and abilities that drive Talos forward. In our pursuit of excellence, we foster a culture of trust, integrity, collaboration, and mutual growth. We are a tight-knit yet globally distributed team of highly experienced engineers and business leaders, with hubs in New York, London, Singapore, and Cyprus. You’ll be part of a hybrid-friendly environment where your unique talents and insights will play a crucial role in building something extraordinary. PRINCIPAL ENGINEER, DATA QUALITY PLATFORM You'll own data correctness across Talos. You'll lead the data quality engineering team, run BAU data operations, and own reference and security master data, along with the pipelines and platform that deliver it. Our goal: the data we return to clients is always correct, and we find defects before a client does. You'll join a small team with the mandate to work independently from product engineering and challenge it. Engineering teams make assumptions about how their systems behave. Your job is to test those assumptions and prove or disprove them with evidence.   RESPONSIBILITIES - Lead AI-first data quality engineering - Validate every client-facing endpoint continuously. The focus is our REST and WebSocket APIs. Cloud delivery (S3, Snowflake, Databricks) is also in scope. Assert correctness against independent sources of truth, across every environment and client entitlement. - Cover streaming as well as request/response. For WebSocket, that means: - sequencing and gap detection - snapshot-to-increment consistency - reconnect and replay correctness - staleness of live subscriptions - agreement between the streamed view and the REST view of the same state - Use adaptive detection, not static thresholds. Learn baselines per endpoint and per market, and let them adjust as "normal" changes. Hand-set limits get noisy, go stale, and teach people to ignore alerts. - Apply LLM and agentic tooling to generate checks from API specs and schemas, keep them current as the product changes, and triage alerts before a human sees them. Know when a deterministic rule, invariant or cross-source reconciliation is the better tool. - Test resilience and load. Design load, performance and failure-mode tests. - When you find a fault, raise it with the owning team with enough evidence for them to act. Drive it to resolution, then add a check that would have caught it earlier. - Measure and publish endpoint and field coverage, time to detect, detected-before-client rate and false-positive rate. - Run data governance - Build and maintain a catalog of our data assets and APIs, covering lineage and downstream consumers. Maintain an understanding of assets, how they're used and what breaks when they change. - Own data governance tooling such as OpenMetadata. - Own reference and security master data - Lead the build-out of the next-generation security master platform, then run it day to day: new listings, symbol mapping across venues, and increment and contract specs. - Design processes for safe updates to security entries, with validation and rollba