Founding Engineer (AI & Agentic Systems)

Codexa is building an agent-first CRM. Instead of reps working a CRM by hand, our AI agents run the revenue operation itself: outbound, qualification, and pipeline management, end-to-end, replacing traditional SDR and AE workflows. The MVP is live, we have paying clients, and the core architecture is already running client campaigns. We're looking to hire a founding engineer to own the intelligence layer of the product. This person lives entirely on the agentic and AI side of Codexa: the orchestration that decides what every agent does next, the LLM systems behind it, and the voice infrastructure that runs live calls. It's an AI-focused seat, but not an AI-only one. We're building a full CRM, so you need real full-stack engineering behind you. Compensation This role starts on a fully deferred cash comp rather than a standard salary. It's a market-rate role with pay deferred and repaid at a premium once we hit a funding or revenue milestone. We're upfront that this is a real trade: less cash now, real upside as the company grows. Happy to walk through the specifics directly. What you'd own The agentic orchestration layer: the state machine and decision engine that runs each lead end-to-end across email and voice, including our Next Best Action logic, escalation, and handoff rules Our LLM stack: model integration and routing, prompt systems, tool and function calling, retrieval and vector memory, and the evals and guardrails that keep agent behavior safe and reliable in production Voice AI: the real-time voice agent stack, from call orchestration and latency tuning to qualification on live calls The reasoning and data enrichment pipeline as it scales from internal tooling into a productized platform Application development around all of it: the backend services, data models, and APIs the agents run on Engineering standards for how we build AI systems from the ground up, where your calls on frameworks and architecture will stick What we're looking for Hands-on experience building agentic systems with orchestration frameworks like LangGraph and LangChain, not just calling an LLM in a loop Strong applied LLM engineering: prompt design, tool and function calling, RAG, vector databases, and structured outputs, with real production experience Voice AI experience: real-time voice agents, telephony, and the ability to structure and tune latency-sensitive call flows Comfort integrating and routing across model providers (OpenAI, Anthropic/Claude, OpenRouter) and reasoning about cost, latency, and reliability trade-offs Solid full-stack engineering: able to ship end-to-end across backend, data, and APIs (Node.js/TypeScript, Python, PostgreSQL, queues and workers) Experience with multi-agent or event-driven architectures and state machines Startup or 0-to-1 experience preferred, and comfortable with ambiguity and fast iteration Based in Canada or the US with overlap in Eastern hours. Strong pluses Experience with the Model Context Protocol (MCP) for giving agents controlled access to internal and external systems LLM observability and evals: tracing, prompt versioning, and offline and online evaluation Multi-tenant, workspace-isolated architecture at scale Experience building enrichment or data pipelines that feed agent decisioning