What I can do for you
Overview
Systems Architect and Applied-AI Researcher specializing in agent safety, runtime containment, and governance-first infrastructure. Designer of production distributed systems serving millions across 100+ countries. Deep expertise in cryptographic protocols, sealed computation, multi-party governance, and verifiable inference. M.S. Mathematics (NYU Courant). Published researcher targeting NeurIPS 2026 and POPL 2027. Teaches applied AI at IE University New York City.Core Strengths
Agent Safety & Containment. Propose-Only Action Layers. M-of-N Auditor Governance. Sealed Computation (AMIs, TPM, Nitro Attestation). Cryptographic Protocols & Verification. Distributed Consensus. End-to-End Encryption. Content/Abuse Classification. Verifiable Inference. Security Auditing & Threat Modeling. Production Systems Architecture. Three major platforms: Safebots (agent safety substrate, 2024+), Intercoin (blockchain governance, 8 EVM mainnets), Qbix (federated social platform, 80 repositories, plugin-based full-stack framework).
Objective
Building the next generation of safe, governance-first AI computing infrastructure. Looking to apply expertise in AI safety mechanisms, sealed computation, orchestration, and distributed systems to help organizations deploy AI at scale with strong security and auditability guarantees. I bring battle-tested systems design, peer-reviewed research on AI safety, and a track record of taking ambitious technical ideas from zero to production while maintaining integrity at scale.Impact & Scale
Delivered apps to millions of users across 100+ countries in 15+ languages, with $0 marketing spend. Authored peer-reviewed research on AI workflow safety, cryptographic protocols, and distributed consensus. Designed systems that power community infrastructure worldwide. Passed independent security audits and built codebases handling sensitive user data across 8+ blockchains and millions of interactions.
AI Safety & Research Focus
AI Safety & Agent Containment
Designed SafeBox: a substrate where LLM tools propose writes only; every side effect executes only after M-of-N auditor approval against a signed capability manifest declared before execution. Runs on sealed AWS AMIs with ZFS, TPM/Nitro attestation, and zero remote-access vectors. Closes the in-permission tool-misuse gap that conventional access controls cannot address. Conducted 9+ security audit rounds, fixing critical governance and execution verification flaws. Analyzed production incidents (deleted databases, ignored safety rules, mass email blasts) and mapped each to propose/approve mitigations. Authored LAWS paper on AI workflow safety (arXiv 2605.04069).Governance & Verification
Architected multi-agent orchestration system (Workflows, Steps, Workloads, Tasks, Tools) with cryptographic execution hashing, capability-matched stream materialization, and M-of-N auditor governance. Every tool call is logged and auditable; all writes route through propose-only action layer. Built Safebots as production orchestration engine earning redeemable credits for execution integrity. Designed and implemented OpenClaiming Protocol with RFC 8785 canonicalization, EIP-712 signing, EIP-2771 forwarding across PHP, JavaScript, and Solidity. Cryptographic verification enables cross-domain action attestation without centralized trust.Privacy & Data Protection
Architected Streams: typed graph database with per-node access control, copy-on-write workspace forks, append-only audit logs, and deduplication via content hashing. Implemented OpenClaiming Protocol with RFC 8785 canonicalization, EIP-712 signing, and EIP-2771 forwarding across PHP, JavaScript, and Solidity. Advocated for end-to-end encryption and cryptographic proof over trust-based infrastructure. Designed data governance frameworks for tracking lineage, enforcing purpose limitation, and managing retention at scale.
Academic Research & Publications
Peer-reviewed research spanning AI safety, inference efficiency, cryptography, and distributed systems. Faculty affiliate at IE University New York City teaching AI and emerging technology.
AI & Applications
SafeBox: Sealed Computation with Warrant-Governed Execution (IEEE S&P 2027, HotCRP #153)
Safety & ML Research
Distributed Systems
U Programming Language (POPL 2027, HTML version)
Leadership & Execution
Team Leadership
Built and scaled engineering teams across startups and ventures. Hired, trained, and mentored developers across multiple timezones; designed systems and delegation models enabling teams to ship at velocity. Conducted technical interviews and capability assessments. Led cross-functional teams from zero to product-market fit. Managed development roadmaps, set technical priorities, and communicated architecture decisions clearly to stakeholders. Maintained high-performance teams through transparent communication and results-oriented compensation models.Technical Strategy & Execution
Took startups from concept to deployment. Designed UI/UX, created interactive prototypes and technical architecture diagrams. Managed legal documentation and third-party security audits. Filed patents. Interfaced with investors, consulted with customers, wrote requirements documents and Statements of Work. Pre-sales and enterprise sales. App store optimization and launch strategy.
Technical Expertise
Systems & Infrastructure
Architected and maintain large production codebases powering social platforms, distributed systems, and AI orchestration. Qbix: an open-source social operating system — a plugin-based full-stack framework (PHP + Node.js) for federated social platforms. Apps built on Qbix reached 5M+ users across 100+ countries in 15+ languages with zero marketing spend. Designed Streams typed graph database with per-node access control, audit logs, and efficient materialization. Built non-interactive end-to-end encryption for iOS (hash-chain key derivation, homeserver-mediated routing, full transcript simulability). Optimized Q.js framework (40KB, zero-dependency, 15 years in production). AWS infrastructure with sealed AMIs, ZFS storage, TPM attestation. Built content/abuse classifiers on noisy social data at scale (Bloomberg autocomplete materially reducing help-desk load, Canvs/Mashwork real-time audience-sentiment for television and marketing). Passed independent security audits and cryptographic audits.Languages & Technologies
Languages: Python, JavaScript/TypeScript, Node.js, PHP, Solidity, Swift/Objective-C, Go, Rust, SQL.AI & ML: LLM application systems (RAG, summarization, speaker diarization, multimodal), content/abuse classifiers, evaluation design, TensorFlow.js, transformer internals (attention, MoE, KV cache), verifiable inference.
Distributed Systems & Crypto: Cryptographic protocols (EIP-712, EIP-2771, RFC 8785), consensus mechanisms, sealed computation, Byzantine fault tolerance, blockchain integration (8+ EVM mainnets).
Infrastructure & Security: AWS (EC2, RDS, Aurora, S3), Docker, CI/CD, end-to-end encryption, TPM/Nitro attestation, threat modeling, security audits, production reliability.
Open-Source Platforms & Repositories
- Safebots AI Orchestration (2025+): Workflow engine with cryptographic execution hashing, capability-matched stream materialization, M-of-N auditor governance. Includes Grokers knowledge graph, Code AI-guided development, and OpenClaiming Protocol for action verification.
- OpenClaiming Protocol (2026): Cross-domain credential system with RFC 8785 canonicalization, EIP-712 signing, EIP-2771 forwarding. Implemented in PHP, JavaScript, and Solidity for universal interoperability.
- Intercoin: Distributed Web3 Infrastructure (2018+): 12 audited smart contract types deployed on 8+ EVM mainnets. Community currency infrastructure with governance and reward mechanisms serving 100K+ users.
- Qbix: Open-Source Social Operating System (2011+): A plugin-based full-stack framework (PHP + Node.js) enabling federated social platforms. Includes Streams (typed graph database with per-node access control, audit logs, typed relationships), Q.js (40KB zero-dependency framework, 15 years in production), and modular plugins for identity, messaging, calendars, places, media, and real-time infrastructure. Apps built on Qbix reached 5M+ downloads across 100+ countries.
Development approach (2025 onwards): Architected entire codebase from ground up; used Claude AI for collaborative expansion and solution building on top of existing frameworks. Pre-2025 projects: Architected solutions; wrote ~50% of code with team. All codebases extensively audited and battle-tested in production.
Videos & Presentations
Screenshots & Photos
Built on Qbix5M+ downloads across 100+ countries, zero marketing spend. Data via AppFigures.
Decentralization Before It Was CoolPhilosophy shaped by conversations on surveillance and privacy, web3 and decentralization, and pioneers like Tim Berners-Lee. Encrypted, distributed systems are not idealistic—they are necessary.
Education
New York University
Master's Degree from NYU Courant Institute of Mathematical Sciences in Applied Mathematics – #1 ranked nationally.Graduated 2006. (GPA 3.8).
Brooklyn College, CUNY
B.S. in Computer Science and Mathematics (dual major), Literature minor.Graduated 2003. (GPA 3.6). Also taught as adjunct professor.
Early Achievements & Foundations
Early College & Teaching
Completed high school at 13 with a Regents Diploma and enrolled in Brooklyn College, where I majored in Mathematics and Computer Science. Later taught as an adjunct professor, which reinforced my passion for clear communication and helping others understand complex technical concepts.
Philosophy & Approach
To Product & Safety
I believe that the software and AI systems we build have profound power to shape society, relationships, and user autonomy. Every system I architect prioritizes user privacy, security, and control. I conduct security audits rigorously and implement cryptographic best practices. I avoid dark patterns and designs that exploit user attention or lock in data.Products should help users accomplish real-world goals and then get out of the way. Access controls must be granular: users should know exactly what notifications they receive and why. Databases should be designed for minimal exposure: end-to-end encryption and cryptographic proof-of-integrity are superior to trust-based infrastructure. I believe regulations like GDPR are best solved through technology, not compliance theater. Massive data breaches demonstrate that centralized data stores are catastrophic risks—distributed, encrypted systems are the answer. See my thoughts on surveillance and decentralization.
To People & Community
I am passionate about building tools that empower people and unite communities. Products must give people control over their data, identity, and relationships—not extract value from them.History shows that collaboration beats competition in the long run. Open-source software, Wikipedia, and peer-reviewed science have created the greatest wealth explosion and technological progress humanity has seen. I have pioneered results-oriented compensation models that align team incentives and reward actual contribution.
I believe the future involves universal basic income, more time with families, and increased participation in open-source projects—software, drug research, scientific analysis, political analysis. The best outcomes come when smart people collaborate toward shared goals without extractive intermediaries. See my thoughts on technology, security, and community.
