WildPath.io is a small independent practice helping companies lead, build, and assess technology. We're hiring for one role.
Open Role: AI Engineer, Agentic Systems & Hardening
The role
Companies are putting LLM agents into production faster than they can secure or operate them. You will design and build agentic systems for clients, and harden the ones they already have, so they are reliable, observable, cost-controlled, and safe to give real permissions to. You'll also contribute to the AI portion of our technical due diligence work, judging whether a target's AI features are a strength or a liability.
What you'll do
- Design and build agentic systems: tool use, multi-step planning, memory, retrieval, and multi-agent orchestration.
- Harden agents against prompt injection, data exfiltration, excessive permissions, and unsafe tool calls, using least-privilege tool scopes, sandboxing, input and output validation, and human-in-the-loop approvals.
- Make agents dependable: evals and regression suites, tracing and observability, retries and fallbacks, timeouts, and failure containment.
- Control cost and vendor risk: token budgets, caching, model routing, and avoiding lock-in to a single model provider.
- Threat-model and review client AI systems, and write up findings in plain language with prioritized fixes.
- Support technical due diligence on AI-heavy targets: architecture, model dependencies, data rights, and cost per customer.
What we're looking for
- Hands-on experience shipping LLM-based agents to production, not only prototypes.
- Strong software engineering fundamentals in Python and/or TypeScript, and comfort with APIs, queues, and cloud infrastructure (AWS, Azure, or GCP).
- A working knowledge of AI security: the OWASP Top 10 for LLM Applications, prompt injection, tool and permission abuse, and secrets handling.
- Experience building evals and monitoring for non-deterministic systems.
- Clear written communication. You can explain a risk to a founder or an investor without jargon.
Nice to have
- Experience with MCP servers, agent frameworks, or durable workflow engines.
- Security background: threat modeling, red-teaming, or application security.
- Consulting, M&A, or due diligence experience.
- FinOps experience with LLM spend.
How we work
- Remote and engagement-based. Work is scoped per client.
- Independence. We say what we find, including when it isn't what the client hoped to hear.
How to apply
Email [email protected] with your LinkedIn profile or CV and a short note on an agentic system you built or hardened: what went wrong, and what you changed. Links to code or write-ups are welcome.
Don't see a fit but want to work with us? Send a note anyway. We read everything.