About the Company
LoneTree Capital is a lower-middle-market private equity and growth equity firm. We back founder-owned companies in durable end markets with mission-critical, recurring revenue, and we partner with management teams to build them into category leaders — growth investing with the discipline of private equity. We invest behind fragmented, slow-to-change markets where operational rigor compounds. Our operating platform, APEX, is how we bring that rigor to portfolio companies. We run a lean, high-output team where individual contributors have outsized impact and direct line of sight to the partners.
About the Role
LoneTree is building AI into how the firm works as part of how the investment team operates day to day. Deal screening, sector research, landscape building, diligence support, LP and fund materials, portfolio monitoring: every one of these is a workflow that an AI engineer can make faster, sharper, and more repeatable. We have already adopted tools like Claude and ChatGPT across the team. What we don’t yet have is someone who builds. This role exists to turn ad-hoc AI usage into durable, production-grade tooling: the agents, skills, and integrations that take real work off the team’s plate and make our process compound. This is a build-from-scratch role, working with the tools we already have. You will define how LoneTree runs on AI. It is not a research role, a strategy role, or a program-management role; you will be expected to ship things, measure whether they worked, and iterate. You will be the technical substance behind LoneTree’s AI capability, reporting directly to the Managing Partner.
Responsibilities
Build the Tooling (60%)
- Go where the work is. Embed with the investment team to learn how the work actually gets done: sourcing, screening, diligence, IC prep, fundraising, portfolio monitoring and find where AI can meaningfully simplify it. The unit of work is real workflows, not novelty demos.
- Ship agents that do real work. Design and ship multi-step, tool-using agents that take work off the team’s plate: deal screeners, sector-research and landscape automations, diligence-question generators, document and data extraction, LP and fund-material drafting. Bias toward production-ready, not demo-ready.
- Build durable, not disposable. The agent you ship should leave behind reusable skills the team can pick up, not a black box only you understand. Think in primitives and patterns, not one-off scripts.
- Take evals and guardrails seriously. You know why a 95%-reliable agent is sometimes worse than a 70%-reliable one with a clear handoff, especially when the output feeds an investment decision, and you build accordingly.
- Instrument what you build. Define the metrics, prove the impact in time saved, errors avoided, or work that no longer needs doing.
Shape the Platform (20%)
- Define how LoneTree’s AI tooling is built, evaluated, deployed, and observed; the patterns the rest of the firm’s AI work compounds against.
- Establish the skills pattern: how reusable capabilities get packaged, versioned, evaluated, and shared. This becomes the unit of leverage across the firm.
- Build the integrations that connect LLMs to the systems we actually use; data providers, CRM, document stores, our research and deal stack via MCP servers and APIs.
- Evaluate and recommend AI tooling: foundation models, agent frameworks, MCP servers, eval and observability tooling. Make clear build-vs-buy-vs-integrate calls.
Enable the Team (15%)
- Hand work back. Train the team on what you ship, document the patterns, and transition ownership so you can move to the next workflow.
- Run working sessions and office hours calibrated to the audience — a partner one day, an analyst the next.
- Translate what you ship into the language of the business. Surface wins. Build the momentum that drives adoption.
Portfolio & Strategy Partnership (5%)
- Where it creates value, extend what you build into portfolio companies as an operating lever and partnering through the APEX platform.
- Partner with the Managing Partner on LoneTree’s AI roadmap; produce the proof points that inform our team and, where relevant, our LP narrative.
- Track the AI model, vendor, and agent-platform landscape and bring informed recommendations to the firm.
What This Role Is Not
- An ML or MLOps engineering role. We are not training models.
- A research role. We apply what exists; we do not publish papers.
- A data-science or quant-research role. You will build tooling for the team, not investment models.
- A program-management role. You will not coordinate other people’s work; you will do the work.
Qualifications
- 8+ years building production software. The bulk of this role is shipping, and you should be comfortable doing it without a team behind you.
- Direct experience building agents multi-step, tool-using LLM workflows running in a production or near-production context. You’ve thought about evals, guardrails, failure modes, cost, and handoff to humans.
- Working knowledge of the modern agent stack: foundation model APIs (Anthropic, OpenAI, or equivalent), at least one agent framework (LangGraph, OpenAI Agents SDK, Anthropic tool use, or similar), MCP, and evaluation tooling.
- Strong fluency across the stack: scripting, light app development, API integrations, data pipelines, evaluation harnesses. You can build the thing, not just specify it.
- Track record of driving adoption of new tools and practices across an organization, with measurable results to show for it.
- Exceptional written and verbal communication. You can explain an eval result to a non-technical partner and debug a tool-call chain in the same afternoon.
- Comfort with ambiguity. You will define the path, not walk one that’s already paved.
