Senior Machine Learning Engineer - AWS, Real-Time Inference, Pipelines
About The Position
At TWG AI, we drive innovation and business transformation across a range of industries—including financial services, insurance, technology, media, and sports—by leveraging data and AI as core assets. Our AI-first, cloud-native approach delivers real-time intelligence and interactive business applications, empowering informed decision-making for both customers and employees. We prioritize responsible data and AI practices, ensuring ethical standards and regulatory compliance. Our decentralized structure enables each business unit to operate autonomously, supported by a central AI Solutions Group, while strategic partnerships with leading data and AI vendors fuel game-changing efforts in marketing, operations, and product development. You will collaborate with management to advance our data and analytics transformation, enhance productivity, and enable agile, data-driven decisions. By leveraging relationships with top tech startups and universities, you will help create competitive advantages and drive enterprise innovation. At TWG, your contributions will support our goal of sustained growth and superior returns, as we deliver rare value and impact across our businesses.
Requirements
- Strong data / ML engineering experience with streaming systems (e.g., Kafka / Kinesis / MSK) and modern data storage formats
- Experience building low-latency, high-throughput inference services
- Proficiency in a systems language (e.g., Go) alongside Python
- Production AWS experience
Nice To Haves
- Familiarity with financial market data or trading protocols a plus — the US feed is a FIX 5.0 SP2 drop-copy session with real-world quirks (nanosecond timestamps, repeating groups, dedup semantics)
- Familiarity with chain-data infrastructure (node providers, subgraphs, event indexing) is a plus
Responsibilities
- Streaming and storage pipelines that feed both model training and low-latency inference.
- The online inference path and its latency. A real-time detector microservice is built and unit-tested but not yet deployed — it needs to be connected to a run-time model and hold latency under live load. One known, non-trivial problem lives here: batch scoring ranks across a whole population, but single-account (or single-wallet) real-time scoring has no population to rank against, so it must threshold on calibrated raw scores.
- The model retraining cadence as data and labels accumulate, including drift-triggered retraining.
- Productionizing new features and detectors on the fast path, in partnership with data science.
Benefits
- A bonus will be provided as part of the compensation package, in addition to a full range of medical, financial, and/or other benefits.
Stand Out From the Crowd
Upload your resume and get instant feedback on how well it matches this job.
What This Job Offers
Job Type
Full-time
Career Level
Senior
Education Level
No Education Listed
