Projects
A living ledger of what I build and why. Each project includes Problem → Actions → Outcomes → Credits → Artifacts → Status (plus quick facts, my role, and next milestones).
A living ledger of what I build and why. Each project includes Problem → Actions → Outcomes → Credits → Artifacts → Status (plus quick facts, my role, and next milestones). This page exists to make the work verifiable, useful, and attributable.
APort and the Open Agent Passport - Live (current work)
Quick Facts
| Scope | Pre-action authorization for autonomous AI agents |
|---|---|
| Stage | Live, open specification and reference implementation |
| ICP | Teams running AI agents that touch money, data, or infrastructure |
| My role | Founder, author of the specification and the paper |
| Highlights | Open Agent Passport spec, MIT licence, DOI 10.5281/zenodo.18901596; 21 published policy packs; 7 framework integrations |
| Core metrics | 53ms median authorization latency; 0 percent attack success under restrictive policy across 800+ attempts |
Problem
An AI agent decides what to do with a probability distribution. A wire transfer, a database drop and a shell command are not probabilistic acts. The gap between what an agent can do and what it should do is an authorization problem, not an alignment problem, and training a model harder does not close it. Neither does reviewing the logs afterwards.
Before APort there was no standard place to put that decision. Teams either trusted the model or wrote ad hoc checks inside their own application code, which an injected prompt can talk its way around.
Actions
- Wrote the Open Agent Passport specification: a signed credential binding an agent's identity to the capabilities and limits it is allowed to exercise.
- Built the enforcement point as a
before_tool_callhook at the framework layer rather than inside the model, so a prompt injection cannot reason its way past it. - Made the decision deterministic. Same inputs, same answer, no sampling and no temperature, terminating in allow, deny, or escalate.
- Published 21 policy packs and integrations for 7 agent frameworks.
- Ran a public adversarial testbed, the APort Vault CTF, with a 5,000 dollar bounty, and published the results.
Outcomes
- 3,900+ authorization decisions across 1,100+ sessions in the live CTF, over half denied.
- over 70 percent attacker success against the permissive tier. 0 percent against the restrictive tier across 800+ attempts. Nobody claimed the bounty.
- Median authorization latency 53ms, p95 63ms, p99 76ms over 1,000 requests.
- Specification released under the MIT licence with a DOI, and mapped to the NIST AI Risk Management Framework as a proposed Pre-Action Authorization control category.
Opportunity created: The result that matters is the boring one. Players social-engineered the model successfully and the infrastructure still refused the transfer. That says the fix is architectural, so a team can adopt it without waiting for a better model.
Artifacts
- Paper: Before the Tool Call: Deterministic Pre-Action Authorization for Autonomous AI Agents
- Specification and reference implementation: github.com/aporthq
- Live testbed and results: vault.aport.io
Status
Live and current. This is what I work on now.
Chimoney - Acquired by CapitalSage Vantage
Quick Facts
| Scope | Payments infrastructure (collections, payouts, multicurrency accounts) |
|---|---|
| Stage | Wound down May 2026, acquisition agreement signed June 2026, closing in phases |
| ICP | Platforms, NGOs, fintechs, global payouts/collections |
| My role | Founder and CEO. Leading the transition through close |
| Highlights | Interledger integrations; CAD/USD/NGN balances; payouts to 100+ countries; email/phone payments; UPA links |
| Core metrics | 80% GM, 99% on-time delivery, 99.9% uptime; top corridors: Nigeria, India, Kenya, Canada |
Problem
Cross-border payouts and collections are fragmented, slow, and expensive. Mid-market orgs need an API-first, compliance-aware layer that “just works” across countries and rails.
Actions
- Built payment orchestration layer (collections, payouts, wallets) with compliance guardrails.
- Shipped multi-currency accounts (CAD, USD, NGN, MXN) and email/phone payouts.
- Integrated Interledger and rolled out Universal Payment (UPA) links.
- Established corridors and local partners to reach 100+ countries.
- Delivered SLAs around uptime, latency, delivery success.
Outcomes
- 99% on-time delivery, 99.9% uptime.
- Live corridors used by NGOs and platforms; top corridors: Nigeria, India, Kenya, Canada.
- Select Clients: Goodwall, Developer Nation, Interledger.
Opportunity created: Enabled NGOs & platforms to reliably pay global contributors/beneficiaries, unlocking work in regions with limited financial access. Reduced time-to-integrate from 4 months to 2 weeks for API and Infrastructure clients.
Credits
Chimoney Team • Early customers (Microsoft Africa Development Center, Goodwall, Developer Nation) • Interledger Foundation & community • Corridor partners • Advisors & investors.
Artifacts
- Website & Docs: chimoney.io • API docs • Public Consumer App info (CAD/USD/NGN balances, UPA, Interledger).
- Press/PR: Chimoney Blog
Status
Live (scaling)
Next Milestones
- Corridors: reliability & FX improvements in top 5 lanes.
- Scale transaction volume
- Interledger GA
World Innovation League (WIL) → DTTP - Live
Quick Facts
| Scope | Training → talent → opportunity; community-to-work pipelines |
|---|---|
| Stage | Operating, expanding |
| Program | DTTP (Diverse Tech Talent Program) |
| My role | Founder — program design, partnerships, funding, reporting |
| Funding | $1.5M raised (WIL) |
| Footprint | Training, mentorship, sprints, hackathons leading to jobs & projects |
Problem
Talented people are under-connected to real work. Organizations lack de-risked pipelines to skilled contributors.
Actions
- Designed DTTP: structured sprints from learning → hackathon → mentorship → work experience → employer match/interviews.
- Ran training-to-job, mentorship, and hackathon-to-work programs; aligned outputs to partner needs.
- Built reporting, compliance, and grant management to keep programs sustainable.
- Raised funding for the program and Cohorts
- Partnered with delivery partners like Co.Lab, Riipen, Atila, SkillHat, Founder Institute, and others.
Outcomes
- $1.5M raised to deliver training/opportunity at scale.
- 800 participants trained; 85% placed/contracted/further training; 100+ active mentors.
Opportunity created: Converted learning into paid, resume-worthy work, especially for underrepresented talent. Scaling access to to real world work experience and Startup creation.
Credits
WIL Core Team • Mentors • Partner orgs & donors • Volunteers • Participants.
Artifacts
- Program Highlight & syllabi • Mentor/Employer Guide • Featured Participant project • Partner Showcase
Status
Live (continuously improving)
Next Milestones
- Launch a Foundational AI Skills Cohort.
- Raise $5m by December 2028.
- Employer-side “brief to bounty” intake to accelerate paid work.
Oruly - Live
0:00
/0:48
1×
Quick Facts
| Scope | Hybrid AI + human workforce: post tasks, build hybrid teams, manage work |
|---|---|
| Audience | Startups, teams, and operators who need tasks done by AI, humans, or both |
| Core value | Post Tasks → Build Hybrid Team → Get Results in one place (tasks, comms, tracking) oruly.ai |
| Pricing notes | Free platform access, pay-as-you-go AI task credits, low team seat pricing oruly.ai |
| My role | Founder — product direction, positioning, integrations |
Problem
Teams bounce between marketplaces, freelancers, and AI tools; coordination costs (handoffs, tracking, QA) kill momentum.
Actions
- Shipped a task marketplace + team builder to combine AI and human workers under one workflow.
- Added project tracking + conversations so task posters see progress end-to-end in Oruly.
- Published transparent, pay-as-you-go pricing (credits for AI tasks, optional seats).
- Added Challenge (Hackathon) management, hosting, and participation using the Team Feature.
Outcomes
- Clear promise and funnel: “Get shit done with AI + Human Collaboration” resonates; users can post tasks or build persistent hybrid teams and track results.
- Used to run team-based events and hackathons like at DTTP Hackathons
- 2000+ active users.
Opportunity created: collaboration layer for AI and Human to complete, lowering context switching.
Credits
Oruly core team, early posters/workers, DTTP Hackathon team.
Artifacts
- Site: oruly.ai (How it works, Features, Pricing, Task Types).
- Oruly Tasks and Jobs
- Oruly Blog
Status
Live (iterating)
Next Milestones
- Deeper AI–human orchestration (QA loops, review workflows).
- More team workflows (retainer tasks, SLAs).
- Expanded task categories + verified worker profiles.
- More AI skills to enable one-click deployment of AI agents in teams
Freelance Ledger - Live
Quick Facts
Problem
Freelancers bleed time on admin: invoices, chasing payments, spreadsheets, tax prep, none of it billable. They also miss out on opportunities to access tax benefits and savings.
Actions
- Built Smart Invoicing (one-click, auto-filled clients, payment links, recurring billing).
- Shipped AI voice input to log income/expenses and even create invoices hands-free.
- Added analytics dashboard (income vs expenses, trends, exportable reports) + tax savings wizard.
- Implemented payment integration (Chimoney) and client portal for a professional client experience.
Outcomes
- A single, focused workspace where freelancers generate invoices in seconds, track payments, and see real-time financial health—reducing admin drag.
- MVP live; Users growing
- Reduced admin time reported by early users.
Opportunity created: Let independents focus on paid work, not paperwork.
Credits
Early adopters and feedback community, payment partners, and contributors to guides.
Artifacts
- Site + Features: freelanceledger.com (Smart Invoicing, Voice Input, Analytics, Tax tools).
- Guides (e.g., How to Invoice Clients).
Status
Live
Next Milestones
- Deeper regional tax automations, payment reminders, and improved desktop and mobile voice logging using GPT-5.
- Payments links, reminders, basic analytics; 10 growth partners.
Food Waste Log - Live
Quick Facts
| Scope | Restaurant food-waste tracking + a growing library of free ops templates (inventory, temperature logs, recipe costing, feedback forms) |
|---|---|
| Core value | Track/measure food waste to cut cost & improve sustainability; provide plug-and-play docs for the back-of-house |
| My role | Creator — product direction and resource curation |
Problem
Restaurants lose margin to untracked pre- and post-consumer waste (spoilage, over-production, misfires, remakes). Without systematic logging, teams can’t see patterns or fix root causes.
Actions
- Built Food Waste Log as a focused hub for waste tracking and restaurant ops resources.
- Published free templates to accelerate adoption:
- Inventory Management Spreadsheet
- Recipe Costing Calculator
- Food Temperature Log
- Customer Feedback Forms (incl. healthcare variant)
Outcomes
- Clear pathway for kitchens to start logging immediately (templates) and standardize processes, aligning with industry guidance that waste logs help pinpoint problems and reduce losses.
- 10 free templates created
Opportunity created: Saves money, reduces waste, and informs smarter shopping/cooking.
Credits
Early restaurant testers, template contributors, and sustainability advisors.
Artifacts
- Site: foodwastelog.com (resources hub + free downloads).
- Free Tools and Templates: https://foodwastelog.com/app/resources
Status
Live (MVP)
Next Milestones
- In-product waste logging with analytics (by station, time, reason).
- POS/Inventory integrations; weekly “Top 5 Waste Items” report.
- AI analysis and personalized recommendations for waste reduction and cost-cutting
AfricaHacks Graduated → WIL
Quick Facts
| Scope | Pan-African hackathons → talent pipelines |
|---|---|
| Stage | Graduated |
| My role | Founder — community building, partnerships |
| Legacy | Evolved into World Innovation League |
Problem
Aspiring builders across Africa had limited access to structured opportunities, mentors, and pathways to work.
Actions
- Ran large-scale hackathons, mentorship, and showcases.
- Partnered with companies to turn winning projects into internships and paid work.
Outcomes
- Built one of the largest tech communities of its kind in Africa ( 30,000+ participants).
- Led to the creation of WIL as the sustained vehicle for training → opportunity.
Opportunity created: Thousands gained access to mentors, tools, and first work wins.
Credits
Volunteers • Mentors • Sponsors • Partner orgs • Alumni.
Artifacts
Status
Graduated (program DNA lives on in WIL/DTTP)
NaijaHacks Graduated → AfricaHacks
Quick Facts
| Scope | National hackathon & community (Nigeria) |
|---|---|
| Stage | Graduated |
| My role | Founder — program design, sponsorships, ops |
| Legacy | Scaled into AfricaHacks, then WIL |
Problem
Local builders needed a first on-ramp to community, mentorship, and real projects.
Actions
- Launched NaijaHacks; created tracks aligned to local needs and partner briefs.
- Scaled playbook regionally → AfricaHacks.
Outcomes
- 5,000+ participants; 50 partners and sponsors.
- Direct lineage into AfricaHacks → WIL.
Opportunity created: First wins for new builders; direct path from event to employment.
Credits
Volunteers • Sponsors • Judges • Mentors • Alumni.
Artifacts
Status
Graduated (superseded by AfricaHacks → WIL)
Notes on Status Labels
- Live — In active use with users/partners.
- In development — Building; limited pilots/betas.
- Graduated — Evolved into a new, larger vehicle.
- Sunsetted — Deliberately ended; learnings retained.
- Exited/Left — Transitioned out; stewardship passed to others.
This page updates as projects evolve. If you benefited from any project and want to add your story or artifact, send a note. I’ll include it so the record stays accurate and useful.
Last updated: September 4, 2026