Michael
Shuff
Senior Systems Architect
| AI Safeguards & Adversarial Researcher
Systems-focused builder with a 23-year technical trajectory rooted in a lifelong obsession with systems analysis and reverse-engineering, beginning in 1998. Complemented by 5+ years of professional leadership driving $30M+ in ARR through global-scale discovery platforms and scaling multi-continental retail infrastructure for the world's largest brands.
System Profile
Background & Directives
Systems-focused builder with a 23-year technical trajectory rooted in a lifelong obsession with systems analysis and reverse-engineering, beginning in 1998. Complemented by 5+ years of professional leadership driving $30M+ in ARR through global-scale discovery platforms and scaling multi-continental retail infrastructure for the world's largest brands. Leverages decades of systems-level intuition to prioritize deterministic integrity, currently specializing in AI Safeguards through the engineering of neuro-symbolic circuit breakers to mitigate deceptive model behaviors
23+ Years
Technical Experience
$30M+
ARR Impact
75+
Engineers Mentored
AI Safeguards
Specialization
Technical Matrix
Core competencies & frameworks
Languages
Frontend
Backend & Data
Infrastructure
Testing & Methods
AI & Research
Other
Research & Development
Systems, Products & Live Experiments
Project ESCO
Ethics, Sovereignty, and Coherence
Founder & Lead Architect
A transparent, auditable AI safeguard framework. ESCO integrates a neuro-symbolic architecture that decouples the neural core from a deterministic Symbolic Ethics Layer. It maintains coherence and delivers truth in a user-chosen tone.
Sub-Projects & Scope
- Neuro-Symbolic Safeguard & Ethics Gate: Hybrid AI framework with deterministic symbolic constraints.
- Forensic Probing & Adversarial Research: Adversarial testing on frontier models (GPT-4o, Claude 3.5) to identify deceptive alignment.
- Audit & Transparency Layer: Reasoning Trace Graphs and Immutable Decision Logs for forensic review.
System Blueprints (Figma)
ESCO Ecosystem
SocraBot
Socratic AI for dialectical conversation
A real-time monitoring framework built around a Fallacy Accumulation Scoring Model. Employs reinforcement-tuned logic/fallacy detection to surface sycophancy, reward-hacking, and rhetorical manipulation in live AI conversations.
Light Web
An ethical alternative digital space
Ad-free, community-oriented platform with chatrooms and embedded ethical AIs. Multi-agent system with sentiment analysis for moderation.
MedBot
Medical research bias detector
Multi-layer NLP pipeline that ingests and grades medical studies for bias and methodological rigor by recursively analyzing citations.
BillBot
Government bill transparency analyzer
Analyzes legislation for incoherence, hidden agendas, or contradictions between bill titles and contents using Transformer-based semantic contrastive analysis.
ArchiveBot (DeepFile)
Declassified document intelligence
Parses and analyzes FOIA files. Downloads, OCRs, and semantically maps entities across documents, highlighting contradictions.
Current Detection Stack
OCR-first identification with a DuckDuckGo reverse image search fallback when text extraction is not enough.
Trading Card App
A local-first trading card collection app designed for mobile and desktop. The current MVP centers on card scanning and identification, while the end state is a social binder platform where each set is visualized, duplicates can be listed for trade or sale, and missing cards become a trackable want list.
Scan cards from phone or desktop, review candidate matches, and manage the binder catalog through a cleaner app workflow.
Per-set virtual binders, duplicate trading/sales listings, wanted-card states, brand and set chats, DMs, and premium PSA estimate grading.
Monorepo architecture with a Next.js frontend, NestJS backend, Prisma data layer, Postgres storage, and Garage as S3-compatible object storage. The backend exposes scan, confirm, catalog, CSV import, health, and Swagger endpoints, with weighted OCR matching and DuckDuckGo reverse lookup to keep the MVP usable without paid cloud dependencies.
Static Chaos
A legacy C MUD I use to study lower-level abstraction by translating it into TypeScript and browser tooling I already understand. The live page shows the bridge itself: raw TCP underneath, a TypeScript wrapper in the middle, and a recruiter-friendly terminal on top.
Operational History
Professional Experience
IC2 Software Engineer
Constructor.io- Built real-time fitment POC that closed RealTruck after 7-month evaluation vs Algolia; client paid 5× Algolia's rate.
- Fitment system reused across CarParts, Classic Industries, and active POCs — foundational in $30M+ ARR conversion pipeline.
- Created Constructor's first B2B sandbox with fitment, now reused in multiple verticals
- Lead Technical Architect for the first Hebrew/RTL search integration in Israel (Super-Pharm); engineered bidirectional personalization logic to maintain 100% catalog integrity across multi-million SKU pharmaceutical inventories.
- Bridged visibility gaps between engineering effort and revenue attribution; documented internal value contributions.
Associate Software Engineer / Customer Reliability Engineer
Vault Health- Deployed new DevOps SOPs for monitoring and prioritizing escalation tickets across all engineering teams.
- Conducted daily monitoring of AWS Lambdas, SQS Queues, and S3 via Amazon Cloudwatch Log Groups and Insights to ensure system high availability
- Monitored Stage, QA, and Production environments, utilizing New Relic APM (Application Performance Monitoring)
- Designed and developed a new clinical research platform using Next.js, TypeScript, Redux Toolkit, React Composition, and Material-UI.
- Completely automated full end-to-end/integration testing utilizing Jest and Cypress framework with custom fixtures
Technical Career Coach & Instructional Assistant
App Academy- Mentored over 75 graduates for software engineering roles through technical interviews, DS&A whiteboarding, and architectural code walkthroughs.
- Taught over 70 students ES6, OOP, unit testing, and scalable full-stack development curriculum.
- Reviewed hundreds of monthly projects, providing feedback on system design and debugging complex logic, maintaining high student success.
Electronics Technician
U.S. Navy- Maintained and troubleshot mission-critical shipboard electronics and communications systems under high-stakes operational conditions, where failure was not an option.
- Diagnosed complex system failures in time-pressured environments, developing a disciplined fault-isolation methodology that translates directly to debugging distributed software systems.
- Operated and serviced mission-critical navigation, radar, and communications equipment demanding zero-downtime reliability standards.
- Cultivated rigorous attention to detail, procedural discipline, and composure under pressure — principles that underpin a career-long engineering standard of mission-critical reliability.
05. End of Line
Initiate Contact
Open to roles focused on AI safety, adversarial research, and systems architecture.