Software Engineer – Agentic Platform

Most engineers who use AI tools are operators: they prompt an agent, review the output, and ship the result. This role is different. You're not just using the pipeline – you're building it.

Nectar and Vivery are platforms that connect people facing food insecurity with community organizations and the wraparound support they need. We're a small team moving fast, and we've built an agentic build pipeline – a system of orchestrated AI coding agents, evaluation harnesses, guardrails, and feedback loops – that lets us build and operate software at a pace a traditional team our size couldn't match.

That pipeline is how we're building Nectar 2.0, an agentic platform where AI is the primary author of new code.

This role works in close collaboration with our CTO, who sets technical direction. You'll extend and strengthen the pipeline itself, build full-stack features across both platforms, and carry real production ownership for what you ship. This is a you-build-it-you-run-it team. The work is underway now, and you'll be stepping into it, not starting from scratch.

If you're the kind of engineer who wants to understand systems at every level, who'd rather design the machinery than just operate it, and who's energized by work that has a direct impact on real people – this is a rare seat.

Nectar connects food-insecure Neighbors with food pantries, nutrition programs, benefits and community-based organizations. The code you ship is the connective tissue between healthcare systems, community food networks, and the neighbors they both serve.

When the pipeline produces bad output and it reaches production, real people are affected. That's not a reason to be slow – it's a reason to build the right guardrails, design for blast radius, and take production ownership seriously. The engineers here care about that standard because of what's on the other end, not because a process document says to.

• Design and maintain the orchestration layer: the pipelines, agent configurations, evaluation harnesses, guardrails, and feedback loops that let AI coding agents produce real, deployable software reliably and at scale.
• Go beyond prompting. This work is about building the machinery around AI agents – the systems that catch failures, validate output, and keep the pipeline trustworthy even when an individual agent isn't.
• Establish and refine the patterns that make a heavily AI-authored codebase something the whole team can build on, review, and trust in production.

• Build and ship end-to-end features across both platforms, from database to API to UI.
• Work across the stack without strong language preference. The right tool for the problem matters more than a favorite language.
• Collaborate closely with the engineering lead on architecture decisions, with the autonomy to drive implementation without the path being laid out step by step.

• Build and maintain CI/CD, infrastructure tooling, and internal automation wherever it removes toil and accelerates the team.
• Treat the pipeline itself as a production system: monitored, instrumented, and owned.
• When something breaks, you're in the rotation. On-call ownership is part of the role for the systems you build.

• Ship solutions that limit blast radius when something goes wrong, rather than chasing perfection that arrives too late.
• Know how to contain a failure, roll back cleanly, and learn from it faster than a traditional team could.
• Bring a pragmatic view of risk: what's safe to ship now, what needs more runway, and why.

• Not an AI prompt engineer. Skilled use of AI coding tools is table stakes here, not the job. The job is building the systems around those tools.
• Not a single-stack specialist. Engineers who want to work in one language or one layer of the stack are not a fit. Full-stack means full-stack.
• Not a feature factory contributor. You're not filling a backlog someone else owns. You're making architectural decisions, carrying production responsibility, and…