AI Engineering Portfolio
Charlie Fuller
Building AI systems that work. Multi-agent platforms, evaluation frameworks, and strategic tooling — designed for production use, built with problem-first methodology, grounded in governance as code.
01
What I build
From enterprise governance platforms to targeted AI tools — every project starts with the problem, not the technology.
AESOP Transformation OS
Enterprise AI Agent Governance & Operationalization OS
A governance-first operating system that takes an enterprise from “we should use AI” to governed, evaluated, deployed, and monitored agents running at portfolio scale. Ten-stage pipeline, 21 specialized agents, seven evaluation modes, and a hard safety/bias veto that makes critical failures unshippable.
Next.js 16FastAPISupabaseClaude APIVoyage AI
1
Discover & Validate — Structured interview, stakeholder map, GO / NO-GO triage
2
Design & Build — PRD, architecture scoring, 6-phase instruction writing with HITL
3
Evaluate & Certify — Seven evaluation modes, weighted scoring, safety/bias veto
4
Operationalize & Monitor — Launch packages, deploy, cost tracking, drift alerts
Client engagement
SuperAssistant
Executive AI Assistant Platform
Multi-tenant AI assistant platform combining chat, voice interviews, and knowledge base integration. Full-stack architecture serving hundreds of users with role-based access and real-time streaming.
Next.jsFastAPISupabaseElevenLabsRailway
See how it works →
Contract Review
AI-Powered Legal Document Analysis
Five specialized agents analyze contracts for risk, compliance, and negotiation leverage. RAG-powered with async parallel processing for complex documents — each finding links back to the specific clause.
Next.jsFastAPICeleryRedisVoyage AI
See how it works →
PuRDy
AI Product Requirements Assistant
Structured PRD creation through guided conversation. Problem-first methodology ensures requirements are grounded in validated user needs before solutions are proposed — it refuses to jump to features until the problem is clear.
Claude APIStructured promptingAnthropic SDK
See how it works →
Waifinder Assistant
AI Career Guidance System
Knowledge-base driven career guidance assistant built from 6 months of practitioner transcripts, academic research, and tactical playbooks. Helps professionals navigate the transition into AI consulting with research-backed, confidence-rated guidance.
Claude APICassidyDeep Research
See how it works →
02
How I work
Principles that hold across every project, regardless of stack or scale.
Principle 01Problem-first,
not technology-first
not technology-first
Validate the problem before a single requirement is written. If the problem isn’t real, costly, solvable, and ours to solve, no amount of technology fixes it.
Principle 02Governance
as code
as code
Ethics, safety, and bias checks aren’t documentation — they’re embedded in the pipeline as hard gates. Critical failures are unshippable by design, not by policy.
Principle 03Full lifecycle
thinking
thinking
Discovery, design, build, evaluation, deployment, repair, and monitoring — all one connected pipeline. A deployed agent without operationalization is a future incident.
Principle 04Human in the loop,
always
always
AI generates, humans approve. Every build pipeline has a checkpoint. Every evaluation shows its work. Every repair is tracked with rationale. Automation accelerates judgment — it doesn’t replace it.
03
Let’s talk
If you’re building AI systems and want to do it right — governed, evaluated, and built to last — reach out.