Forward Deployed Engineer

Build the AI systems that change how a portfolio of software businesses operates. Own a whole business function, from the first conversation with the people doing the work to a system they rely on every day.

LEO Software Group buys and operates software businesses for the long term. We are part of Vela and Constellation Software. We are building a team to redesign core business functions around what AI can do now. Each engineer will own one function across multiple portfolio businesses.

Why This Role

You will work closely with our AI Services Lead and our internal teams to turn an ambitious AI strategy into working systems. You will help shape each solution, own the engineering from discovery through production, and see its effect on the business.

The ambition is to build systems that understand the business, act across its tools, and improve through evidence from real work. That means connecting fragmented data to business meaning, giving agents reliable ways to act, and closing the loop between execution, evaluation, and improvement. You will shape how this works across businesses with different CRMs and legacy software.

What Will You Do

• Find the simplest path to the outcome. Work with the team to understand the problem and establish a baseline. Remove unnecessary steps before automating. Build a first working version within days and use feedback to decide what deserves further investment.
• Give agents an accurate understanding of the business. Connect operational systems, documents, and conversations to the context each task needs. Resolve identities and conflicting definitions; make data freshness, provenance, and permissions explicit. Design ontologies, semantic layers, and knowledge graphs where they improve understanding and execution.
• Engineer agents that finish the work. Build or adapt agent harnesses for tasks that span multiple steps, tools, and sessions. Design context and memory, durable state, recovery, and clear completion criteria. Make consequential actions controlled and observable, with human intervention where needed.
• Make evaluation part of development. Build evals from real tasks with the people who know what good looks like. Assess outputs and resulting system state, inspect execution traces, and turn production failures into regression cases. Use that evidence to improve data, tools, prompts, models, and orchestration.
• Take the system into production. Write and ship the software yourself using current AI development tools. Own integrations, access controls, monitoring, and failure recovery. Make explicit tradeoffs between task success, cost, latency, and operational risk.
• Own adoption and business results. Work alongside users, teach the new workflow, and keep improving it until their numbers move. A technically impressive agent that nobody uses, or that leaves the underlying work unchanged, is unfinished.
• Make each deployment compound. Turn working patterns into reusable software that adapts across portfolio businesses. Adding another business must not require another FDE. Once a function runs reliably with monitoring and maintenance, hand it over and rebuild the next one.

What Success Looks Like

Within your first six months, your function has a working system in production, people using it, and a measurable improvement against an agreed baseline. You have taken it beyond the first business, with subsequent deployments getting faster and no matching increase in FDE headcount.

What You Bring

• You build systems people use. You have personally taken software from an ambiguous problem to production, owned the engineering yourself, and delivered a meaningful result. At least one system you have built with current AI tools is live and used. Private work counts equally.
• You actively explore what AI can do. You build and experiment without waiting for an assignment. Recent models, tools, and techniques have changed how you work, and you have formed your own views about where they succeed and fail.
• You have depth in applied AI engineering. You have shipped systems that connect models to business data and tools, evaluate performance on real tasks, and operate reliably in production. You understand context, retrieval, agent orchestration, and failure recovery. You can judge when an ontology, knowledge graph, or custom harness improves the solution—and when a simpler approach is better.
• You own the business outcome. You understand how people actually work, identify what can be removed, and build around the result they need. You take responsibility for solving the problem while challenging assumptions about the solution. Your work continues through adoption and measurable improvement.
• You operate independently and work well with others. You turn unclear requirements into working software, get yourself unblocked, and make sound tradeoffs. You earn users’ trust, explain decisions clearly, and use feedback to improve the product. Once a system can run without your constant attention, you are ready to build the next one.

We care about exceptional work and what it accomplished. You might have built and led an internal AI product, shipped a product people depend on, or founded and grown a software business. What matters is what you personally built, the decisions you owned, and the results you delivered. There is no degree or years-of-experience requirement.

How We Hire

After reviewing your application, we invite selected candidates directly to a paid working trial lasting a few hours. Using sanitized business data and the AI tools you normally use, you will build a working proof of concept for an agentic system that carries out real work. We want to see how far you can take a real problem in that time, with a scope that reflects the job you would be doing.

You will also meet members of the team you would work with. We confirm the time commitment and payment before you start.

For each, explain what you built, who uses it, and the result. Include what you are building now and a recent AI development that changed how you work.