Software Development Engineer 3- AI & Agentic Systems
Wex · Portland, ME ·
- Employment
- Full time
- Category
- Software engineering
Wex · Portland, ME ·
At WEX, a Software Development Engineer (SDE) Level 3 is a senior technical leader who designs complete services, solves complex integration problems, and drives the lifecycle of major initiatives from end to end.
In this role, you will directly participate in WEX’s fast-paced AI-driven opportunities, acting as a key builder and strategist for next-generation technology. You will champion internal AI adoption by helping build AI-powered sofrware (using tools such as Google Gemini, AWS Agent Core, Claude, and other third-party agentic integrations) while sharing education and coaching within the organization.
What You’ll Do
AI & Agentic Engineering
Agentic Workflows: Design and deploy multi-agent workflows within AWS Agent Core, effectively handling state management, memory persistence, and complex tool interactions.
AI Leadership & Enablement: Help drive internal education and coaching initiatives for tools with as Google Gemini Enterprise; collaborate with external AI partners on new agentic opportunities.
Architecture & System Design
End-to-End Ownership: Design, build, and own the lifecycle of entire services and cross-platform integrations. Solve complex system gaps and drive overarching audit strategies.
Strategic Data Orchestration: Make critical trade-off decisions on data orchestration (e.g., MCP/API, Agents/Skills) based on robust data and requirements.
Resilient Design: Build systems that handle failure gracefully by implementing retries, circuit breakers, and fallbacks. Ensure all designs are testable and maintainable long-term.
Security First: Review code for security flaws and ensure all services adhere strictly to WEX’s authentication standards (e.g., OAuth scopes).
Cloud Infrastructure, Reliability & CI/CD
Cost-Aware Cloud Architecture: Design reliable cloud infrastructure (e.g., auto-scaling groups, failover strategies) while proactively avoiding resource over-provisioning.
Deployment & Environment Management: Own the automated deployment package and metadata framework. Write scripts or automation manifests to deploy configurations cleanly across sandbox tiers, preventing environment drift.
CI/CD Optimization: Build and optimize CI/CD pipelines, focusing heavily on reducing build times and tightening feedback loops.
Root Cause Resolution: Capable of solving complex bugs (intermittent issues, race conditions, memory leaks), ensuring the root cause is fixed permanently.
Incident & Operations Leadership: Own Incidents for your services during outages. Create step-by-step triage runbooks and playbooks that empower Tier 1/2 helpdesk teams or junior analysts to resolve common platform errors without escalation.
Product Partnership & Technical Leadership
Requirement Definition: Partner with Product Managers who define Functional requirements by owning the Non-Functional requirements (Performance, Security, Scalability) needed to meet business KPIs.
Backlog & Scope Negotiation: Work with PMs to prioritize the backlog, explicitly negotiating the inclusion of "Tech Debt" and "Refactoring" by explaining long-term business risks. Negotiate "Must Haves" vs. "Nice to Haves" to ensure sprint achievability.
Solution-Oriented Mindset: You don't say "No"; you say "Not like that." You offer alternative solutions that respect time and budget constraints.
Empathic Mentorship & Review: Enforce strict quality standards during code reviews with empathy ("Check tone"). Reject sloppy or brittle "quick fixes" while setting the organizational standard for clean design patterns. Proactively mentor SDE 1s and 2s, delegating work specifically to help them grow.
Consensus Building: Facilitate data-driven discussions to resolve engineering disagreements, effectively de-escalating "religious wars" about technology and ensuring "quiet" voices are heard in design meetings.
Continuous Learning: Rapidly learn and apply new languages or frameworks, lead knowledge-sharing sessions, seek out critical feedback, and context-switch effectively across workstreams when project parameters change.