Why USAA?
At USAA, our mission is to empower our members to achieve financial security through highly competitive products, exceptional service and trusted advice. We seek to be the #1 choice for the military community and their families.
Embrace a fulfilling career at USAA, where our core values – honesty, integrity, loyalty and service – define how we treat each other and our members. Be part of what truly makes us special and impactful.
We are proud to support active-duty military spouses. USAA roles may offer remote or hybrid flexibility for active-duty military spouses consistent with applicable policy and business needs.
The Opportunity
The Application AI Ops team is where machine learning meets the business logic customers touch — the services, APIs, and transaction flows that move money and keep accounts current.
As a Senior AI Ops Engineer on this team, you won’t be prototyping in a notebook and handing it off for someone else to ship: you’ll build the ingestion pipelines that turn application logs, traces, and business-transaction signals into features; train and tune the models that catch a degrading payment API or an anomalous checkout failure before a customer calls in; and then write the deterministic, tested automation that acts on what those models find — a guided remediation, a safe auto-restart, a rollback that actually rolls back. This is a from-scratch build inside a regulated financial institution, which means your code will eventually touch production systems that move real money, under real audit scrutiny — and the engineering bar reflects that.
This position offers a flexible work environment with an onsite requirement of 5 days per week. This position is based in San Antonio, TX. Relocation assistance is not available for this position.
What you'll do:
Build and operate high-throughput ingestion and feature pipelines that consume application logs, traces, APM data, and business-transaction events — collected by the Observability & Telemetry team — into the AIOps feature store.
Train, validate, and tune anomaly-detection and forecasting models against application and transaction-level signals, and build LLM-based log parsing, summarization, and root-cause-hypothesis components with explicit handling for hallucination risk.
Design and build event correlation and probable-cause ranking logic that reduces raw application alert volume reaching a human.
Write, test, and version remediation automation for application-layer failure modes, with tested rollback for every action, operating within the approval-gate, blast-radius, and kill-switch framework.
Instrument and monitor your own models and automations in production, retraining on a defined cadence, and participate in the on-call rotation for any live action at autonomy level 3 or above.
undefined
What you have:
Bachelor's degree; OR 4 years of relevant education and/or experience.
5+ years of hands-on software or data engineering experience, with strong programming ability in Python plus a second production language — code that ships, is tested, and is code-reviewed, not just notebooks.
Practical experience building and productionizing ML models — anomaly detection, classification, or forecasting — using frameworks such as TensorFlow, PyTorch, or scikit-learn, including the full lifecycle from training through retraining.
Working fluency with LLM-based techniques — prompt design, retrieval-augmented generation, structured outputs — and a clear-eyed understanding of their failure modes.
Experience writing and testing operational automation — workflow orchestration, runbook-as-code — including rollback design and approval-gated or policy-controlled execution.
Comfort operating in a regulated environment, with demonstrated ability to debug and reason about a live production incident under time pressure.
What sets you apart:
Proven experience deploying model-driven root-cause analysis or incident triage solutions used in production application…