Felipe F.

Most AI agents die in the demo. I build the ones that survive production.

I shipped a deterministic AI agent into a regulated B2B workflow at a US enterprise SaaS company: schema-bound tool calls, validation at every boundary, human confirmation before execution. Client setup went from days to under 15 minutes. Zero unsafe tool calls since launch.

I also wrote publicly about what did not work, and why we changed direction. That honesty is the job.

What I do:
→ Take an LLM feature from "cool demo" to "safe in production"
→ Design guardrails: tool schemas, validation layers, confirmation gates, failure modes
→ Scope AI features with real technical feasibility, not wishful thinking
→ Audit an existing agent and tell you why it hallucinates

What I bring beyond AI:
12+ years in product across enterprise SaaS, a Tier 1 telecom with 60M subscribers, and a streaming platform where I ran a zero downtime migration of 335,000 subscribers.

I work async, I write clearly, and I ship.

Based in Sydney, available for US and EU handover workflows. Case studies in my portfolio below.