Building AI agents and the infrastructure they run on.

I work at the intersection of agentic systems, financial AI, and regulatory governance — where the gap between what AI can do and what institutions can trust is still wide open. I build AI systems that are architecturally precise, auditable, and honest about uncertainty.

Published research: 100+ citations across ethical AI in financial decisioning, credit risk modeling, supply chain finance, and systemic risk.

→ omoshola.me

• 🏗️ I design complex AI systems as composed layers: data contracts, model services, reasoning engines, policy controls, and observable execution paths.
• 🔁 I prioritize architecture-level guarantees: traceability, determinism where required, graceful degradation, and explicit failure boundaries.
• 🧭 I treat governance as a system primitive, not an afterthought: explainability, audit logs, access control, and policy enforcement are built into core workflows.
• ⚙️ I enjoy hard systems problems at scale: multi-agent orchestration, graph-native memory, temporal reasoning, and reliability under real-world constraints.

• 🦀 Nexus — Supply chain intelligence platform in Rust. Treats demand, lead times, and supplier reliability as probability distributions. Monte Carlo simulation, temporal graph kernel, agentic recommendations a procurement team can audit.
• 💳 FyxCred — Credit intelligence built on behavioral cashflow data, not bureau scores. Three scores — financial health, income stability, financial resilience — with a consent-first data layer and bias transparency framework.
• 🧱 ZEX — Multi-product workspace and execution layer coordinating core platform workflows across the ZEX portfolio.
• ✅ Verity — Verity backend implementation for truth-state validation and system-of-record outcomes.
• 🚆 ZEXRail — ZexRail backend implementation for rail operations and domain workflow orchestration.
• 💸 ZEXPay — ZexPay backend product workflow, integrated to create Verity proposals and consume Verity outcomes.
• 🧠 AgenticMemory — Persistent memory for AI agents. A binary graph format that stores facts, decisions, corrections, and reasoning chains with semantic edge types.
• 👁️ AgenticVision — Persistent visual memory for AI agents with embeddings, similarity search, and visual diff.
• 🧩 AgenticCodebase — Semantic code intelligence for concept mapping, grounding, and impact analysis.
• 🪪 AgenticIdentity — Cryptographic identity, trust delegation, and signed action receipts for agents.
• ⏱️ AgenticTime — Temporal reasoning for schedules, deadlines, decay models, and sequence-aware context.
• 📜 AgenticContract — Policy contracts and governance controls for obligations, approvals, and violations.
• 📡 AgenticComm — Structured communication layer for channels, pub/sub routing, and agent coordination.

• 🏦 Explainable credit risk intelligence — Architected an explainable ML credit risk pipeline for 200,000+ applications using ensemble modeling + real-time macroeconomic feature integration, with transparent reasoning outputs for underwriting; reduced default rates by 15% while expanding access to underserved populations.
• 🔐 Privacy-preserving synthetic data — Built a synthetic data generation stack combining Gaussian Copula and GAN-based synthesis for secure AI model development, achieving full anonymization with regulatory alignment while preserving statistical utility for downstream risk prediction.
• 📦 Supply chain resilience modeling — Developed forecasting and risk pipelines using ARIMA, neural time-series models, and ML supplier-risk scoring to detect disruptions early; reduced stockouts by 22% and drove $2M in cost savings via model-informed operations.

• 🔬 Explainable Credit Intelligence — SHAP-based interpretable risk scoring (CrossSHAP methodology)
• 🌐 Network Analysis — Supply Chain — Systemic risk assessment in supply chains
• 🏦 Credit Risk Creditum — Credit risk modeling and analysis
• 📊 RiskX — Risk analytics engine
• 📄 AgenticMemory Paper I — Binary format and memory substrate design
• 📄 AgenticMemory Paper II — Query expansion and retrieval strategy
• 📄 AgenticMemory Paper III — MCP server architecture and tooling
• 📄 AgenticVision Paper I — Visual/web graph representation and reasoning
• 📄 AgenticVision Paper II — Vision MCP server design
• 📄 AgenticCodebase Paper — Semantic compiler and code intelligence architecture
• 📄 AgenticIdentity Paper — Trust anchors, identity proofs, and receipt model
• 📄 AgenticTime Paper — Temporal format and reasoning model
• 📄 AgenticContract Paper — Policy engine and governance constraints
• 📄 AgenticComm Paper — Agent communication runtime and protocol design
• 📄 IEEE Academic Reviewer — Peer reviewer for ICAD 2025
• 📐 IEEE Standards Development — AI ethics, cybersecurity, financial LLM requirements, supply chain security

• 🏅 IEEE Senior Member, 2025 — Elevated in recognition of significant contributions to the profession; eligible for executive volunteer positions and review panel service. Active memberships: IEEE Computational Intelligence Society, IEEE Consumer Technology Society, IEEE Technology and Engineering Management Society, IEEE Young Professionals.
• 🎯 Ethics and Conference Reviewing — NeurIPS 2025 (Datasets & Benchmarks), DeepLearningIndaba 2025, IEEE ICMI 2026 (King Faisal University), IEOM 2025 World Congress (University of Windsor), IEEE IATMSI, and 2025 International Conference on Signal Processing, Computation, Electronics, Power and Telecommunication.
• ⚖️ Judging and Evaluation — TrackShift Innovation Challenge 2025 (Mphasis F1 Foundation x MoneyGram Haas F1 Team), HackNC 2025 (UNC Chapel Hill), ASA Statistics Project Competition (Grades 7-12; 30+ projects), and ASA USCLAP.
• 🧪 Journal Peer Review — Journal of Data Analysis and Information Processing (JDAIP), including LLM-powered enterprise intelligence, healthcare big data, and cloud optimization work.
• 🤝 Mentorship — SciPy Conference 2025 Mentorship Program and Nova Talent Elite Mentorship Program, supporting emerging and senior professionals in ethical AI, financial AI, and supply chain analytics.

• 🏛️ OSTP Federal AI Policy Contributor (2025) — Submitted technical recommendations on AI regulatory reform to address SR 11-7/SR 23-4 barriers to explainable financial AI, including SHAP/LIME recognition, federal data API clarification, and interagency coordination; quantified a 10-15% potential credit-loss reduction impact across the $18.04T household debt market. (Notice 90 FR 46422 • Docket OSTP-TECH-2025-0067 • Submission)
• 🌐 U.S. Commerce Federal AI Export Strategy Contributor (2025) — Submitted strategy guidance for the American AI Exports Program focused on compliance-native, explainable, and secure AI exports (Basel III/IV, GDPR alignment, SHAP/LIME, IEEE harmonization), with emphasis on financial resilience and allied infrastructure development. (Notice 90 FR 48726 • Docket ITA-2025-0070 • Submission)

• 🦀 Nexus — Shipping a supply chain OS in Rust with probabilistic planning, simulation-first decision support, and auditable agent recommendations.
• 💳 FyxCred — Building cashflow-native credit intelligence for credit-invisible populations with explainable scoring, policy-aware decisions, and governance-ready outputs.
• 🧠 Agentra Sisters — Advancing MCP-native, artifact-portable infrastructure across graph memory, multimodal vision, semantic code intelligence, identity, time, policy, and communication.
• 🧪 Applied AI R&D — Developing explainability, uncertainty-aware modeling, and privacy-preserving data methods for regulated production systems.
• 📝 Research Service — Reviewing AI and cybersecurity research through IEEE and related scholarly programs.

• How to engineer trustworthy agent systems over time, not just at launch.
• How to operationalize SHAP/LIME-style explainability so adverse-action reasons satisfy ECOA/FCRA requirements in real workflows.
• How to build credit models that expand access for people historically excluded by formal scoring systems.
• How to combine uncertainty quantification, calibration, and drift monitoring into continuous model governance.
• What memory, identity, policy, and reasoning primitives are required before agents can safely operate in regulated domains.
• How to use privacy-preserving synthetic data to balance data utility, security, and compliance.
• How African knowledge systems can inform modern computation and AI system design.

• Modeling: explainable ML, ensemble risk models, time-series forecasting (ARIMA + neural), supplier-risk scoring.
• AI Safety & Governance: model transparency, adverse-action traceability, policy-aware decision systems, regulatory-compliant AI deployment.
• Data & Privacy: synthetic data generation (Gaussian Copula + GAN), anonymization, utility-preserving data pipelines.
• Agent Infrastructure: graph-native memory, multimodal retrieval, identity and trust primitives, temporal reasoning, contract-constrained actions.