Craig Stueber
Applied AI Engineer
Builds and ships production LLM systems end to end. Doctoral researcher in AI safety.
“Work hard and be nice to people.”
experience
Work History
Senior Full Stack Engineer
AI Systems Integration- Tech lead and people lead for a team of 6 engineers building Dekaflow 2.0, a next-generation enterprise platform supporting billions in annual east coast energy movement.
- Led early-stage AI agent R&D designing a six-agent LangGraph pipeline for enterprise data understanding, translating complex operational data into actionable business insights.
- Built and owned full-stack features end to end across Next.js, Java, MongoDB, and Azure including gas flow scheduling, hourly quantity tracking, and a cross-cutting user preferences system.
- Led enterprise-wide GitHub Copilot deployment across 200+ engineers, establishing behavioral guardrails and governance practices for safe AI adoption.
Senior Full Stack Developer
LLM-Integrated Systems- Sole engineer across 6 independent brand teams, building all customer-facing applications from 0 to 1 without product management support.
- Built an agentic customer service tool combining customer context with rep input to generate complete, ready-to-send response emails with recipient routing and CC recommendations.
- Replaced third-party tooling with in-house solutions, reducing external vendor costs by $250,000 annually.
- Ran controlled prompt A/B evaluations analyzing token sensitivity and regression risks across model versions prior to production rollout.
Co-Founder & Lead Engineer
- Designed an agentic onboarding system using CrewAI with constrained generation patterns to maintain consistent, safe outputs in a consumer-facing context.
- Owned full product architecture and technical direction for a cross-platform React Native and Supabase mobile application.
- Built real-time chat, event scheduling, and location-aware discovery across iOS and Android.
- Led full App Store and Google Play submission including TestFlight and Play Console policy compliance.
Full Stack Engineer
ML-Enhanced IoT Systems- Integrated ML models for time-series anomaly detection and classification into backend services to identify sensor abnormalities and operational risks.
- Built real-time IoT monitoring dashboards using React, Python, and WebSockets, translating ML outputs into actionable insights for field operators.
Earlier Experience
2017 – 2021Frontend delivery across React, PHP, Shopify, and early AR prototypes. Managed requirements and delivery timelines directly with clients.
Frontend modernization, email system consolidation, mobile UX improvements, and accessibility remediation.
Full WCAG and ADA audit and remediation, Wix platform migration, and staff accessibility training.
Full-stack delivery for nonprofits, authors, and real estate clients. Owned requirements, scoping, and delivery without project management support.
projects
Notable Work
CodeRisk Advisor
AI Safety / LLM SystemsMulti-agent AI security review system for Python, JavaScript, and TypeScript code. Combines OWASP Top 10 vulnerability scanning with AI-specific behavioral risk detection using a panel of specialized LLM agents that synthesize findings into conversational developer guidance.
- LangGraph pipeline orchestrating five specialized agents: VulnScanner, BehavioralRisk, Skeptic, Remediation, Synthesizer
- Skeptic agent actively disputes low-confidence findings to reduce false positives
- Token-by-token SSE streaming with real-time agent status updates in the UI
- Deployed on Google Cloud Run with LangSmith tracing for full observability
Dekaflow 2.0
High-stakes enterprise platform managing natural gas scheduling workflows supporting billions in annual east coast energy movement. Built on a modern React and cloud stack integrating with a 25-year-old Java and SQL legacy system.
DanceCard
Agentic onboarding system via CrewAI paired with a full cross-platform React Native social application. Owned all architecture, data modeling, and delivery independently.
Hot Tomato Summer
Multi-city restaurant voting platform reaching 30,000+ users in two weeks with rule-based fraud detection and voting anomaly dashboards.
PurrQuest
Location-aware mobile app for tracking outdoor and stray cats with clean geospatial state management and secure photo uploads.
skills
Technical Skills
Languages
AI & LLM Systems
Frameworks & Libraries
Infrastructure & Cloud
Data & Backend
Testing & Quality
Accessibility
Enterprise Tooling
research
Doctoral Research
Evaluating the Security of AI-Generated Code: A Quantitative Study Using a Custom Scoring Framework
Designs and validates a reproducible hybrid vulnerability scoring framework to detect and measure security risks in AI-generated code before deployment. Addresses a validated gap in the literature -- no systematic evaluation framework existed for assessing AI-generated code security across diverse programming tasks and contexts.
writings
Published Work
The Comfortable Apocalypse
When Survival Isn't the Problem — Irrelevance Is
The central risk of the AI age is not domination or rebellion, but displacement. As automation removes friction from daily life, it quietly erodes the cognitive and emotional capacities that effort once built — memory, judgment, curiosity, creativity, identity, and agency. The danger is not hostile AI, but a world where thinking becomes optional and human participation fades without resistance.
education
Academic Background
Doctor of Philosophy
Computer Science- AI safety and behavioral reliability
- Security risks in AI-generated code
- Hybrid vulnerability scoring framework combining OWASP, CVSS, and AI-specific pattern detection
Master of Science
Information Technology- IT management and information security management
- System design and architecture