Qa Ai Autom

About the Role

Build and own QA and evaluation systems for LLM-powered assistant and API surfaces in a regulated wealth-management platform. Design and operate multi-agent CI workflows that detect, triage, and optionally fix AI-generated defects while enforcing code quality for agent-authored code.

Explicitly requires vibe coding skills — building and evaluating multi-agent AI workflows, prompt engineering, and AI-assisted code generation.

Location: Saint Petersburg, FL 33701

Type: Full-time

Posted: 3 days ago

Category: AI/ML & Data

Company Rating: 3.3 (3 reviews)

Vibe Markers: 💻 Open Source, 🔥 Hot

Experience Level: Senior

Job Description

Role

This is a hands-on engineering role to evaluate, test, and gate non-deterministic LLM-driven systems within a regulated wealth management platform. You will own the AI evaluation framework, wire evaluations into CI, and build multi-agent workflows that detect and remediate defects automatically while scaling code quality controls for agent-authored code.

Key Responsibilities

• Own and extend the AI evaluation framework for chat and API surfaces: correctness, grounding/citation fidelity, retrieval quality, refusal/escalation behavior, tone, latency, and cost per interaction.
• Curate and version golden datasets and adversarial test sets representing real advisor workflows.
• Design LLM-as-judge and rubric-based scoring for non-deterministic outputs and validate judges against human-labeled data.
• Integrate evaluations into CI to gate changes in prompts, models, retrieval, and tools; instrument production traces and close the loop from live failures back to evals.
• Design and operate a multi-agent quality step in CI/CD that executes suites, triages failures, writes defects with reproduction detail, proposes or applies fixes, and re-runs verification.
• Define agent roles, hand-off logic, guardrails, and trust thresholds for auto-applying fixes versus routing to humans; make cost-aware decisions about LLM use vs deterministic automation.
• Define and implement automated gates for agent-authored code: coverage and mutation tests, static analysis, security and dependency scanning, architectural conformance, and review checklists.
• Identify failure modes specific to agent-generated code and build detection for them; partner with architects and delivery teams to maintain velocity without sacrificing quality.
• Maintain and modernize existing automation (API, UI, integration) and mentor engineers on quality practices.

Requirements

Required

• 5+ years in software quality, test automation, or SDET work with ownership of automation architecture.
• Production experience with LLM-powered systems: agent patterns, prompt engineering, tool/function calling, and orchestration.
• Hands-on experience with major model APIs (OpenAI, Anthropic, Azure OpenAI) and AI-assisted development tooling (Claude Code, Copilot, Codex or equivalent).
• Strong programming in at least one: Python, TypeScript/JavaScript, Java, Kotlin, or C#; comfortable reading others.
• Experience building and interpreting evaluation criteria for systems without a single correct answer.
• Deep CI/CD fluency: pipeline integration, gating, monitoring, and logging.
• API testing experience and validating third-party integrations.
• Clear written communication to explain quality risk and root causes across stakeholders.

Preferred

• Experience with eval and observability tooling such as Langfuse, Promptfoo, OpenTelemetry, New Relic, or equivalents.
• Familiarity with RAG fundamentals: embeddings, chunking, vector search, and retrieval evaluation.
• Experience with workflow orchestration tooling (n8n or similar).
• Exposure to Azure cloud services and the .NET ecosystem.
• Experience with BDD/Selenium or comparable UI automation at scale.
• Background in financial services or other regulated domains and experience mentoring or leading distributed QA engineers.

First 90 Days

• Days 1–30: Learn the assistant, API layer, and current eval/automation estate; ship a baseline eval suite for a high-traffic chat workflow with CI-visible results.
• Days 31–60: Stand up the first multi-agent defect-detection step on a single service; establish trust threshold for auto-applied fixes and publish metrics.
• Days 61–90: Propose and begin rolling out agent-authored code quality standards with automated gates across delivery teams.

Tech Stack

OpenAI, Anthropic, Azure OpenAI, Claude Code, Copilot, Codex, Python, TypeScript, JavaScript, Java, Kotlin, C#, Langfuse, Promptfoo, OpenTelemetry, New Relic, n8n, Azure, .NET, BDD, Selenium

Skills

Test Automation, Automation Architecture, Quality Engineering, SDET practices, Evaluation of non-deterministic systems, Prompt Engineering, Agent Orchestration, CI/CD integration, Monitoring & Observability, API Testing, Regression Testing, Root Cause Analysis, Static Analysis & Security Scanning, Data Governance & Compliance, Mentoring & Communication, Triage and Defect Reporting, Designing Evaluation Criteria, Metrics and Instrumentation