Senior AI Engineer - Model Selection and Orchestration

Senior AI Engineer with experience building production AI systems, model orchestration, tool calling, retrieval, and modern software engineering practices to build reliable AI capabilities across customer-facing and internal experiences.

Hands-on engineer who owns components of the decision and orchestration layer that determines when agents should use a model, tool, retrieval system, or deterministic workflow, balancing quality, safety, latency, reliability, and cost.

Remote-first opportunity for US-based employees with the option to work in-person out of our Manhattan office.

Start your adventure with Zip

Become a part of Zip's Engineering team and take on exciting challenges at the intersection of AI, software engineering, and financial technology. At Zip, our engineers build systems that move real money and serve millions of customers, making reliability, security, and sound engineering judgment fundamental to how we build.

We are seeking a Senior AI Engineer to help turn rapidly evolving AI capabilities into dependable customer experiences and useful tools for our teams. You’ll work within our AI Platforms team, building shared capabilities that support Zia, our customer-facing AI experience, as well as internal agents and AI-powered workflows across Zip.

You’ll play a hands-on role in designing and building the decision and orchestration layer that helps agents interpret requests and determine the right execution approach. Depending on the task, that could mean deterministic logic, retrieval, a lightweight classifier, an AI model, a tool call, or a more complex reasoning workflow. You’ll use evidence about task quality, safety, latency, reliability, and cost to recommend and implement those decisions.

This is a builder’s role where you’ll write production code, evaluate AI system behavior, integrate models and tools, contribute to technical decisions, and work closely with Principal and Staff Engineers, Product, domain engineering, Security, and Risk. You’ll use managed capabilities within the Gemini Enterprise Agentic Platform (GEAP) and approved cloud tooling, while building Zip-specific orchestration, decision logic, and integrations where needed.

Interesting problems you will get to solve

Build Intelligent Decision and Routing Systems: Design and implement the decision layer that determines how an agent should respond to a task, including when to use deterministic logic, retrieval, smaller models, reasoning models, tools, or human escalation.

Build Intent and Clarification Flows: Develop intent classification and clarification capabilities that distinguish information requests from actions, identify ambiguous requests, evaluate confidence, and determine when additional information or escalation is required.

Evaluate and Select Models: Benchmark rules, retrieval approaches, classifiers, smaller models, and reasoning models against representative tasks. Use measurable outcomes such as task success, quality, latency, safety, and total inference cost to recommend model-selection decisions.

Build Reliable Model Orchestration: Implement and improve model adapters, routing logic, structured outputs, timeouts, retry strategies, fallback behavior, and observable failure handling. Build orchestration that is testable and avoids unnecessary dependency on any single model provider.

Integrate Models, APIs, and Tools Safely: Connect agents to approved APIs and tools with appropriate authentication, authorization, policy checks, and customer confirmation. Design for duplicate requests, idempotency, recovery, and other failure scenarios when agents perform state-changing actions.

Build Secure AI Systems: Partner with Security and Risk to implement approved input and output protections, sensitive-data handling, prompt-injection defenses, and controlled access. Maintain clear separation between model judgment and the permissions required to execute actions.

Build Reusable AI Platform Capabilities: Develop reusable orchestration patterns, model adapters, routing components, and tooling that make it easier for domain engineering teams and agent builders to create reliable AI-powered experiences.

Measure and Improve AI in Production: Partner with evaluation engineering to test changes before release, monitor system behavior, investigate failures, and continuously improve model selection and orchestration using production evidence.

Build with AI: Use AI-assisted development tools as a core part of your engineering workflow to accelerate development, troubleshoot problems, improve code quality, and write and maintain tests. Apply sound engineering judgment when evaluating AI-generated solutions for correctness, security, performance, and maintainability.

What you will bring to the team

Educational Background: Minimum of 10 years of related experience with a Bachelor’s degree; or 6 years and a Master’s degree; or a PhD with 3 years experience; or equivalent experience.

Software Engineering Experience: 10+ years of professional experience building, testing, deploying, and operating production software, including hands-on delivery of AI or ML-enabled applications beyond prototypes.

Python or TypeScript: Strong experience with Python or TypeScript and modern software engineering practices, including API design, automated testing, distributed services, and debugging across application, model, and tool boundaries.

Production AI Engineering: Practical experience building applications using LLM tool calling, structured outputs, retrieval-augmented generation, and agent or workflow orchestration.

Model Evaluation and Selection: Ability to evaluate classification quality, uncertainty, fallback strategies, and different model or system approaches using measurable outcomes such as task success, latency, reliability, and total inference cost.

Engineering Craftsmanship: Experience with cloud deployment, automated testing, CI/CD, observability, identity and access controls, secrets management, retries, idempotency, and failure recovery in production systems.

AI Engineering Judgment: Strong judgment about when a problem should be solved with deterministic software, retrieval, a model, or a combination of approaches, as well as when an agent should ask for clarification or hand work to a person.

Ownership and Collaboration: Ability to independently deliver complex technical work within shared architecture standards while collaborating effectively with Principal and Staff Engineers, Product, domain engineering teams, Security, and Risk. You communicate tradeoffs clearly, contribute thoughtful code reviews, and help strengthen engineering practices across the team.

AI-Enabled Engineering: Hands-on experience using AI-assisted development tools as part of your day-to-day engineering workflow to develop, debug, test, evaluate, and improve software while applying sound engineering judgment to AI-generated solutions.

Bonus Points: Experience with Google Cloud, Google ADK, Model Garden, or equivalent managed AI platforms; semantic routing; open-weight models; Model Context Protocol (MCP); reusable agent-development frameworks; Azure integrations; or payments, fintech, or other environments where correctness, auditability, and controlled access are essential.

What you will get in return

Zip is a place where you'll get out what you put in. The newness of our sector means we need to move at pace and embrace change, and our promise to you when you join the team is that you'll feel empowered and trusted to make big things happen quickly.

We want you to feel welcome and supported to be yourself and care for yourself at work. You'll have opportunities to grow your skills and your career, surrounded by smart, friendly people and leaders who have your back.

Flexible working culture and incentive programs

Unlimited PTO, generous paid parental leave and leading family support policies

Company-sponsored 401k match