SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Glia is the #1 Banking AI platform, trusted by 700+ banks and credit unions. Team Vector builds customer-facing voice AI agents that handle real conversations for financial institutions—from balance checks to loan inquiries—with industry-leading latency and safety guarantees.
In this role, you will design, build, and ship the core agentic voice AI framework, including conversation orchestration, tool calling, and runtime architecture. You'll develop the harness controls (guardrails, grounding, policy enforcement, verification layers) that make LLM agents trustworthy for banking's regulated environment. You'll work end-to-end on latency optimization across the real-time voice pipeline, unify runtime architecture to reduce configuration friction, and maintain production stability while building next-generation capabilities.
You'll collaborate with product, design, and engineering teams across Europe and Vancouver, turning ambiguous problems into shipped outcomes. The tech stack includes Python, Elixir, TypeScript/React, PostgreSQL, DynamoDB, AWS, Kubernetes, Kafka, and AI tools like Gemini and Claude.
Key aspects of the work: voice AI with sub-millisecond latency constraints; LLM architecture under real regulatory and safety constraints; real impact at scale across hundreds of financial institutions; a global, distributed team that values craft and velocity; daily use of AI-assisted development tools.
REQUIREMENTS:
- Core requirement: Production AI systems experience with deep understanding of LLM architecture (prompting, context management, tool use, agentic flows, retrieval, grounding, evaluation, latency/cost/quality tradeoffs). Real-time or voice AI experience strongly preferred; agent SDK experience (Claude Agent SDK, OpenAI Agents SDK) a plus.
- Architecture & technical judgment: Own design of complex components, make sound tradeoffs, spot risk early.
- Trusted, fast delivery: Ship reliably without cutting corners; track record of de-risking projects and owning quality, including verification of agent-generated code.
- Applied AI fluency: Actively use Claude or similar tools to design, build, review, and ship; direct agents and verify output.
- Proactive: Identify problems before assignment; partner with PMs from idea to delivery; raise issues and drive solutions.
- Adaptable: Thrive in fast-paced environment with shifting priorities and evolving AI landscape.
- Curious: High learning velocity; pick up new tools and practices quickly; share knowledge.
- Cross-continental collaboration: Clear written and verbal communication; reliable overlap of at least 2 hours daily with Pacific Time (9:00–11:00 AM PT) for Vancouver team collaboration.
- Leadership instincts: Set direction, unblock people, raise the bar. Prior tech lead, team lead, or mentorship experience ideal.
