Principal AI-Native Engineer

Most companies are running AI pilots. We are rebuilding how our companies actually operate. Ionic Partners is a holding company with a portfolio of operating businesses. We are not adding AI features to products or rolling assistants out to employees. We are rebuilding the operating systems of entire business functions — engineering, support, finance, marketing, revenue operations — as agentic systems that do the work, in production, every day. We are looking for Principal AI-Native Engineers to build them. This is not a plan we are pitching. Agents already write and review code inside our engineering pipeline. Our own orchestration framework is what those systems run on, and it is on its second generation because the first one taught us what breaks. The knowledge our agents work from is maintained as structured, machine-readable artifacts rather than scraped from wikis, because we learned early that context quality sets the ceiling on everything else. Every hard-won lesson in that sentence cost us something. You will inherit all of it, and the next set of lessons is yours to learn. This is a central team, deliberately flat, that builds agentic systems for every company in the portfolio. There are no internal layers, no leads, and no junior tier. Everyone on it is a principal-level builder reporting directly to the executive who owns the mandate. Scope is concentrated rather than distributed: each engineer carries an unusually large share of the outcome and an unusually direct line to the decisions that matter. You will not be starting from scratch. We already run a production agentic platform in-house — our own orchestration framework, machine-readable context artifacts that give agents grounded knowledge of our products and customers, and automated development and QA agents working inside our engineering pipeline. Your first job is to ramp into a live system that real businesses depend on. Your ongoing job is to extend it, harden it, and use it to make each new function agentic faster than the last. The team is domain-agnostic by design. Subject-matter experts inside each business define what their function needs. You design and build the systems that deliver it. You do not need to be an expert in finance, support, or revenue operations. You need to be exceptional at turning a well-specified problem into a reliable agentic system that runs without you standing over it. And you own what you build. Not a prototype handed to another team. Not a proof of concept that dies in a demo. You ship it, you run it, you watch its evals, you carry its cost line, you improve it. If it breaks, it is yours. If it transforms how a company works, that is yours too. This is not an applied research role, and it is not an AI enablement or transformation role. Your work product is not a strategy, a recommendation, or a rollout plan — it is a working system in production, operating a real business function, measured on whether that function got materially better. The distance between an impressive agentic prototype and a system a business will actually depend on is where this entire job lives: the eval harnesses, the failure modes nobody anticipated, the guardrails, the traces, the fallback paths, the cost per run at volume, the second year of maintenance. It is also not a management role, and it does not become one. There is no reporting line beneath you and no expectation that you build one. Influence here comes from what you ship and the standard you set for the engineers and operators around you. We built the team flat specifically so we would never have to ask an elite builder to stop building to advance. The bar is high, and we intend to keep it there. Each hire is a large fraction of this team, and we would rather stay understaffed than lower it. If you want to build agentic systems that run real businesses — and stay technical while doing it — we would like to hear from you.

Responsibilities *Partner with subject-matter experts across the portfolio companies to turn functional specs into agentic system designs, pressure-testing scope and surfacing failure modes before build. *Design agent architectures: agent boundaries, orchestration and control flow, tool surfaces, state and memory, human-in-the-loop placement, and the split between model-driven and deterministic logic. *Build, ship, and operate production agentic systems that run real business functions end to end. *Extend and harden the in-house orchestration framework — execution control, integrations, permissioning, observability, failure, and recovery behavior — so each build raises the floor for the next. *Improve the automated development and QA agents already running inside our engineering pipeline. *Create and maintain machine-readable context artifacts covering products, customers, processes, and operational data, and keep them current as the businesses change. *Design retrieval and context-assembly strategies, manage context budgets, and treat prompting as versioned, tested, measured engineering. *Integrate agentic systems with live business platforms — product codebases, CRMs, ERPs, support desks, data warehouses, and internal services — including auth, permissioning, rate and cost limits, idempotency, and blast-radius controls. *Define success criteria and build eval harnesses, baselines, regression suites, and human review sampling where automated scoring is insufficient. *Instrument systems with tracing and observability that make agent behavior diagnosable in production. *Build guardrails, fallbacks, retries, circuit breakers, approval gates, audit trails, and rollback paths proportional to what a system can affect. *Track and control token and tool cost per unit of work, and design for unit economics that hold at full volume rather than pilot volume. *Monitor eval trendlines, cost curves, and failure patterns for systems in production, and act on what they show. *Investigate and resolve regressions, incidents, and quality drift in systems you own. *Evolve deployed systems as the business processes they serve change. *Ramp quickly into large, unfamiliar production codebases across our companies and extend them safely rather than working around them. *Extract reusable components, patterns, and abstractions from solved problems so recurring classes of work get cheaper across functions and companies. *Evaluate emerging models, frameworks, agentic techniques, and tooling against real workloads, and make adoption calls based on measured results rather than capability claims. *Define and apply practices for responsible autonomous operation: data exposure, IP protection, security boundaries, and human review requirements. *Review agentic work built elsewhere in the organization and raise the technical standard through direct engagement and demonstrated results. *Work alongside engineers and operators whose day-to-day work these systems change, explaining behavior, limits, and intent. *Retire systems and approaches that measurement shows are not delivering, and document why.

Requirements *Bachelor's degree or higher in Computer Science, Computer Engineering, Software Engineering, or a related field, or equivalent practical experience. We weigh demonstrated engineering work far more heavily than credentials; a strong body of shipped systems fully substitutes for a degree. *8+ years of hands-on software engineering experience building and operating production systems. *3+ years building AI-Native or agentic systems that reached production and were used for real work — not prototypes, internal demos, hackathon projects, or evaluations that stopped at a pilot. *Demonstrated end-to-end ownership: systems you designed, built, shipped, and then operated and improved over time, including responsibility for their failures. *Experience integrating AI systems with the platforms a business actually runs on — production codebases, CRMs, ERPs, support and ticketing systems, data warehouses, internal services — including authentication, permissioning, and controls on what an autonomous system is allowed to do. *Experience designing and running evaluations for non-deterministic systems: defining correctness criteria, building eval harnesses, establishing baselines, and detecting regressions. *Experience operating LLM-based systems in production, including tracing, failure diagnosis, guardrails, and cost management at volume. *Experience in becoming productive quickly inside large, complex codebases you did not write, and extending them safely. *Experience working from specifications or requirements set by domain experts outside your own area of expertise. *Experience with platform, infrastructure, or developer-tooling work — building the systems that other engineers or systems depend on. *People management experience is not required and is not an advantage. This is a principal-level individual contributor role with no reporting line, and we are explicitly open to engineers who have deliberately stayed technical. *Ownership. You treat what you ship as yours indefinitely — the outcomes, the failures, the cost, and the maintenance. You do not look for the point where responsibility transfers to someone else. *Autonomy. You operate from intent rather than instruction, set your own sequence, and make progress in genuine ambiguity without waiting for the picture to resolve. *Intellectual honesty. You report what the data shows, including when it undermines your own work. You retire your own systems when they stop earning their keep and say plainly what did not work. *Judgment. You weigh speed against reliability, autonomy against safety, and elegance against maintainability, and you consistently choose well without a rule to follow. *Adaptability. You expect your techniques to be obsolete within a year and treat that as normal rather than destabilizing. *Collaboration across expertise boundaries. You work well with domain experts who know their function far better than you do, take their specs seriously, and push back with substance when a spec is wrong. *Communication. You explain complex system behavior clearly to technical and non-technical audiences, and you write well enough that your reasoning survives without you in the room. *Curiosity with discipline. You stay current on a fast-moving field and test new capabilities against real workloads before believing it. *High standards, low ceremony. You hold a high bar for yourself and the people around you without needing a process to enforce it.

Nice to have *Experience building agentic systems in enterprise environments, with the constraints that imply: legacy systems, compliance requirements, security review, real data sensitivity, and organizational change. *Experience making a non-engineering business function agentic — support, finance, marketing, revenue operations, or similar. *Experience building shared platform or framework layers that multiple downstream systems depend on. *Experience with enterprise SaaS, ERP systems, or mission-critical business applications. *Experience applying agentic development or QA workflows inside a real engineering pipeline. *Experience modernizing or extending complex legacy systems using AI-assisted approaches. *Experience working across multiple companies, business units, or product lines rather than a single product. *Experience in a fully remote, globally distributed organization. *Open-source contributions, technical writing, or public work on agentic systems.

Benefits We don't call them perks; they're part of what makes working here great. *Access to frontier models, agentic tooling, and infrastructure without procurement friction, plus the authority to evaluate, choose, and replace what the team builds on. *A live production agentic platform to inherit and extend, rather than a blank page — and the mandate to make every company in the portfolio run on what you build. *Unusual breadth. Because you work across a portfolio rather than a single product, you will touch more distinct problem domains in a year here than in almost any single-company role, and see your work compound across all of them. *A principal-level individual contributor path with real scope, real autonomy, and no expectation that you move into management to advance. *We are 100% remote and global. Live your best life wherever that may be, and never lose out on career opportunities because of it. *Flexible work hours. We work asynchronously and don't care when you're online, just that you deliver great results and are there for our customers. *We are dedicated to your growth with consistent and meaningful feedback, support in achieving your personal career goals, and access to leading-edge tools, playbooks, and technology to amplify your experience. *Introductions to thought leaders in the space and webinars on cutting-edge tech hot topics. *Stipend to help set up your ideal home office. *Focus on culture: coffee chats, happy hours, cooking classes, book clubs, and more!