Backend Engineer, AI

Founding Backend / AI Engineer

Remote | Early-stage startup | Equity-oriented opportunity

Zavvis is building a financial observability layer that gives corporate finance teams continuous visibility into their financial operations.

Inspired by how engineering observability transformed monitoring of complex systems, Zavvis applies similar principles to financial data to help finance teams understand what changed, why it changed, what requires attention, and how to move from detection into governed resolution.

We already have a working product, backend architecture, data foundation, and agentic capabilities in production. We are now looking for a strong Backend / AI Engineer to join the founding team and help us harden the existing system and build the next generation of agent-driven workflows.

This role sits directly between backend engineering and applied AI.

We are looking for someone who is comfortable moving across APIs, data contracts, persistence, orchestration, agent/tool integrations, and production reliability—not someone who only works on prompts or isolated models.

What you’ll work on

• Building and extending production backend services for Zavvis’s agentic financial workflows
• Hardening the existing multi-agent architecture and improving reliability across agent-to-agent and agent-to-tool interactions
• Building backend capabilities that support Detect, Explain, Conversation, Decision Steward, monitoring, and future automation workflows
• Designing and implementing APIs, service contracts, state transitions, persistence models, and event-driven workflows
• Integrating LLM-powered components with deterministic backend logic, financial data, policies, and governed business rules
• Improving tool/function calling, structured outputs, context handling, orchestration, and multi-turn agent behavior
• Building reliable stateful workflows that can move from detection and investigation into decision, action, verification, and follow-up
• Implementing background jobs, asynchronous workflows, event handling, retries, idempotency, and failure recovery
• Building approval, authorization, audit, and workflow-control mechanisms for sensitive financial actions
• Integrating backend services with financial systems and external APIs
• Supporting verification and reconciliation workflows so system actions can be independently confirmed
• Improving testing and evaluation across multi-step backend and AI workflows
• Debugging complex production issues across the AI, backend, data, and application layers
• Working closely with the founder, AI/data engineers, and frontend team to ship quickly

What we’re looking for

• Strong Python experience
• Strong backend engineering skills and experience building production APIs and services
• Experience with FastAPI, Flask, Django, or similar backend frameworks
• Hands-on experience building applications with LLMs or agentic AI
• Experience integrating LLMs with tools, APIs, structured data, and application workflows
• Familiarity with tool/function calling, structured outputs, orchestration, and multi-step agent systems
• Experience designing persistent workflows, state machines, background jobs, or event-driven systems
• Strong understanding of REST APIs, databases, authentication, authorization, and backend architecture
• Experience with SQL and relational databases
• Strong debugging and systems-thinking skills
• Comfortable reasoning across multiple layers of a production system
• Comfortable working in a fast-moving early-stage environment

Experience with LangGraph, LangChain, Pydantic, PostgreSQL, Azure, Azure OpenAI, Docker, queues/workers, event-driven architecture, accounting systems, fintech, or financial applications is a plus, but not required.

We care much more about what you can build, debug, and ship than titles or years of experience.

The profile we are looking for

The ideal person is not purely a backend engineer and not purely an AI engineer.

You should be comfortable moving between questions such as:

• How should this workflow persist and recover state?
• Which decisions belong in deterministic backend logic versus the model?
• How should an agent invoke tools safely and predictably?
• How do we make a multi-step workflow idempotent and auditable?
• How should services exchange structured financial context?
• What happens if an external API times out halfway through a workflow?
• How do we prevent duplicate actions or inconsistent state?
• How do we verify that an action actually occurred?
• How do we make an AI-assisted workflow fail safely when evidence is incomplete?
• How do we test the complete workflow instead of only individual functions?

And then be able to open the codebase and implement the answer.

The next phase of Zavvis requires backend systems that can support durable case state, governed decisions, deterministic lifecycle transitions, approvals, external-system integrations, verification, auditability, and reliable agent orchestration. The architecture deliberately keeps financial authority and state-changing actions governed and deterministic rather than handing those responsibilities to an LLM. Zavvis_C10_02_Flagship_Missing_…

The role will also involve building reusable backend infrastructure for controlled actions, asynchronous execution, verification, and multi-step workflows across the product. Zavvis_C10_02_Flagship_Missing_…

Stage & opportunity

Zavvis is currently bootstrapped and preparing for pilots while raising outside capital. We are not yet in a position to offer a conventional salary for this role.

We are looking for someone interested in joining at the founding-team stage, with an ownership/equity-oriented structure and the intention to transition to paid compensation as funding or customer revenue comes in.

We would use the first month as an intensive working period to evaluate execution, ownership, technical judgment, and long-term fit, with the opportunity to agree on a meaningful ongoing role and equity participation.

We are moving quickly and want someone who can join the current sprint, work across the existing backend and AI stack, and help us reach production-grade pilot readiness before the end of Q4.

If interested, send me a message with your background.