Origin

Operation Parallel Canada was the first public manifestation of the Parallel Canada civic intelligence practice. It was built in under 24 hours, solo, as a submission to the Agency 2026 hackathon hosted by the Government of Alberta in Ottawa, April 29, 2026 in front of federal and provincial ministers, deputy ministers, and frontier AI companies. All findings are sourced from publicly available government data. All methodology is open.

AGENCY 2026 · OTTAWA · APRIL 29, 2026
Parallel Canada

The average Canadian household is paying $160 a year into 73 organizations no one was checking.

We pointed AI at every grant the Government of Canada has ever published, and it found 73 organizations that share the structural pattern of a shell. The methodology is open so you can check our work.

Built by Matthew Kirubakaran in a personal capacity as a private citizen. Not affiliated with any employer or government body.

Canada’s public funding data exists. It has never been joined simultaneously across federal and provincial governments until now. Parallel Canada surfaces the patterns no single regulator can currently see.

Built for the people whose job is to ask hard questions about public money.
851,300 ENTITIES SCANNED  ·  3 OPEN DATASETS  ·  7 SIGNALS  ·  OPEN METHODOLOGY

WATCH THE INTRODUCTION

One person. Less than 24 hours. This is what your people could build if you let them.
Non-partisan. Impartial. Open methodology. Built for governments that want to lead on transparency, not react to it.
0
Entities Scanned
0
Entities Flagged
$0.0B
Combined Exposure
$0
Per Canadian Household

A framework for national accountability.

Non-partisan. Impartial. Open methodology.
NON-PARTISAN
The Shell Index scores every entity by the same seven signals regardless of political affiliation, denomination, or organizational type. A conservative-aligned charity and a progressive-aligned charity with identical funding patterns receive identical scores. The methodology has no political opinion. Proactive transparency is not a risk. It is the strongest available defence against the kind of accountability crisis that emerges when patterns go unexamined.
IMPARTIAL
Every flagged entity requires human verification before any conclusion is drawn. No entity has been named publicly without a documented structural explanation. The operating principle of this entire project is four words: patterns, not accusations.
OPEN METHODOLOGY
Every signal, every weight, and every matching rule is documented and replicable. Any journalist, auditor, researcher, or oversight body can independently verify the analysis. The methodology is available for institutional review on request. Transparency is not a feature of this project. It is the foundation.
This project was built independently by a private citizen using publicly available data. It does not represent the views of any employer, government body, or political organization. Participation in Agency 2026 was in a personal capacity only.

Three datasets. Zero shared view. Until now.

The CRA holds two decades of charity filings. Treasury Board publishes every federal grant and contribution. Alberta posts every provincial procurement. All public. All siloed.

Each dataset on its own answers narrow questions: how much did this charity spend on fundraising? Which department signed this agreement? Who won this contract? None of them answer the question that matters: which entities are stacking funding from multiple governments simultaneously, with the same shape that historically precedes accountability failures?

Parallel Canada is the first tool to join them. One entity identifier, three data sources, one composite score: the Shell Index.

How it works

Three steps. One transparent pipeline.

01
SCAN
Pull every CRA charity filing, every federal grant and contribution agreement, and every Alberta procurement record from the last 20 years. Normalize names, addresses, and entity identifiers across all three sources.
02
SCORE
For each of the 851,300 entities, compute the Shell Index, a 0 to 100 composite of seven structural signals including government revenue concentration, cross-jurisdictional stacking, and reporting opacity. One formula, applied identically to every entity.
03
SURFACE
Rank by Shell Index. The 73 entities scoring above 75 are flagged for review. Each carries a status: CLEARED with statutory or quasi-governmental rationale, or UNDER REVIEW pending further work. Director can drill into any entity, consult the agent team, and submit a brief.
Who it is for

Built for the people accountability depends on.

GOVERNMENT LEADERS
Proactive transparency.
Surface cross-jurisdictional funding patterns before they become audit findings or media stories. Use Parallel Canada to demonstrate proactive accountability on your terms.
OVERSIGHT
JOURNALISTS AND RESEARCHERS
Find the story faster.
851,300 entities. Scored. Ranked. Cross-referenced across CRA filings, federal grants, and provincial procurement. Methodology and findings available for review on request.
INVESTIGATION
AUDIT AND OVERSIGHT BODIES
Triage at national scale.
Parallel Canada does not replace audit. It tells auditors where to look. The Shell Index prioritizes 73 entities for human review across two governments simultaneously.
TRIAGE
Parallel Canada is a professional tool for institutional users. It is not a citizen-facing spending dashboard. Every finding requires human verification before any public attribution.
Try it yourself

The Director’s Interface

A self-directed tabletop intelligence exercise. You are the Director. Four AI agents brief you in real time. You investigate. You decide. The findings are yours to act on.

ACT ONE
You receive the brief.
A classified transmission arrives. The National Accountability Task Force has identified a gap in Canada's public funding oversight infrastructure. No single government agency can see across all three datasets simultaneously. You have been chosen to change that. You have three hours.
MISSION BRIEF
ACT TWO
Your team investigates.
Four bespoke AI agents (AXON, CIPHER, GHOST, and COUNSEL) brief you on what they find across 851,300 entities. You examine the signals. You consult each expert. You ask the hard questions. The map lights up. The numbers tell a story. You decide what it means.
LIVE AI INVESTIGATION
ACT THREE
The brief lands.
Your team synthesizes. A Director's Brief is generated with actionable insights and clear next steps. AXON walks you through the findings. You submit to the Minister. The call comes in. The operation is complete. What you do with the findings, that part is up to you.
DIRECTOR'S BRIEF

The full experience takes approximately 5 minutes. No installation. No account. Just open the link.

LAUNCH THE DEMO →
What we found

Four entities, four different stories.

Three of these clear once you read the underlying mandate. One doesn’t. The Shell Index isn’t the verdict; it tells you where to look.

All entities are identified using publicly available data only. No entity is named publicly without human verification and a documented structural explanation. The tool identifies where to look, never what to conclude.

Sherwood Park, Alberta
Alberta Conservation Association
SHELL INDEX
98
Gov revenue
95%
Federal
$19.6M
Department
ECCC
Provincial
$5.6M
CLEAREDStatutory levy
Calgary, Alberta
Calgary Homeless Foundation
SHELL INDEX
84
Gov revenue
94%
Federal
$244.0M
Department
Infrastructure Canada
Provincial
$367.0M
CLEAREDPass-through funder
Calgary, Alberta
National accessArts Centre
SHELL INDEX
86
Gov revenue
59%
Federal
$10.2M
Department
Canadian Heritage
Provincial
$1.3M
UNDER REVIEWUnder review
Ottawa, Ontario
Rideau Hall Foundation
SHELL INDEX
77
Gov revenue
35%
Federal
$75.0M
Department
Canadian Heritage
Provincial
-
CLEAREDQuasi-governmental mandate
National scale
Across all 851,300 entities, 73 carry a Shell Index of 75 or higher. Combined exposure: roughly $2.4 billion in concurrent multi-jurisdiction funding. About $160 per Canadian household. None of these numbers required a single Access-to-Information request. Every input is already public.
69 entities remain unreviewed.
The four entities verified in this analysis all cleared with documented structural explanations. That is the methodology working correctly. But 69 entities above threshold have never been examined cross-jurisdictionally. No single Canadian government currently has the combined view that would make that review possible. Parallel Canada provides that view. The review itself requires human investigators, journalists, and oversight bodies with the mandate to act on what they find.
If the system can correctly clear an entity scoring 98 out of 100 with a documented statutory explanation, there is reason to trust the model when it flags cases that cannot be cleared. The four cleared entities here are not exceptions to the methodology. They are evidence that it works.
The complete dataset of 73 flagged entities is available to accredited journalists, researchers, and government oversight bodies on request. Contact us below.
Methodology

The Parallel Canada Framework

Parallel Canada is not just a tool - it is a portable analytical framework. Point it at any structured government dataset, define the signals that matter, and it produces a scored, ranked, human-reviewable accountability brief. The Shell Index is one instantiation. The same framework applied to healthcare procurement, infrastructure spending, or municipal grants produces equally actionable output.

The methodology is open. Every signal, every weight, and every matching rule is documented and replicable by any researcher, journalist, or oversight body. The analytical framework is available for institutional review on request. The code base is proprietary pending institutional partnership discussions.

Shell Index breakdown · 7 signals
Government revenue concentration
30%
Share of revenue from public sources, multi-year average.
Circular gifting loop membership
20%
Funding flowing through related charities back to associated entities.
Multi-jurisdictional presence
20%
Federal and provincial funding stacked on the same entity.
Federal multi-department capture
10%
Same entity drawing from three or more federal departments.
Amendment creep
10%
Recurring upward agreement amendments above the original ceiling.
Shared address clustering
5%
Multiple funded entities registered to the same physical address.
Post-funding inactivity
5%
Drop-off in reported activity after major disbursements close.
The team

The Intelligence Team

Four bespoke AI agents, each engineered to the depth of a subject matter expert. Together they cover financial analysis, human intelligence, legal methodology, and cross-jurisdictional mapping. Deployed in 24 hours. Available on demand.

These are not chatbots. Each agent was engineered with a distinct knowledge base, a specific analytical scope, a defined voice register, and a mandate scoped to their domain. CIPHER alone replicates the function of a senior financial analyst. COUNSEL replicates a legal advisor. Together they form a cross-functional intelligence team.
AXON
AXON
Lead / Cartographer
AI-powered conversational agent with deep domain expertise. Synthesizes the team, reads the map, and writes the briefing. AXON is the agent who hands you the call.
CIPHER
CIPHER
Financial Analyst
AI-powered conversational agent with deep domain expertise. Reads CRA filings and federal G&C agreements; watches concentration ratios, related-party transactions, and 20-year funding curves.
GHOST
GHOST
Human Intelligence
AI-powered conversational agent with deep domain expertise. Maps the people behind the entity: directors, board overlap, cross-organization ties, and the networks the line items don't show.
COUNSEL
COUNSEL
Legal / Methodology
AI-powered conversational agent with deep domain expertise. Holds the methodology line. We score patterns, not people. We surface signals, not accusations. Clears entities when the mandate explains the shape.
YOU
The Director
The human in the loop. You issue the brief. Your team investigates. The findings are yours to act on.
Matthew Kirubakaran
Creator and Developer
Builder, Concerned Citizen
Over twenty years shipping products across games, platforms, enterprise software, and public sector technology. Always at the intersection of technology, people, and things that actually matter. Parallel Canada was built in less than 24 hours, solo, using publicly available data and tools anyone can access. Not because it was easy. Because it needed to exist.
Participating in a personal capacity. All findings, code, and opinions are his own. Not affiliated with or representing any employer or government body.
“One person. Less than 24 hours. This is what your people could build if you let them.”
Matthew Kirubakaran, Agency 2026, Ottawa, April 29 2026

Parallel Canada was built by one person, in less than 24 hours, using publicly available data and tools anyone can access.

Imagine what becomes possible when the talented, motivated, mission-driven people already inside government are given the tools, the time, and the encouragement to try. Not as a special program. Not as an innovation lab. Just as part of how work gets done.

This hackathon created those conditions for one day. Agency 2026 showed what that looks like in practice. The hope is that it does not stop here. If this project inspires even one team inside government to ask a hard question with a new tool, it will have done more than win a hackathon.

Materials

Get the materials.

Available after April 29, 2026. Enter your email below to be notified when materials are published.

Pitch deck
Investor and funder narrative
COMING SOON
Product requirements document
Scope, users, decisions
COMING SOON
Technical brief
Architecture and stack
COMING SOON
Methodology paper
Shell Index in detail
COMING SOON
FAQ

Common questions.

Credits

Built by Matthew Kirubakaran.

Designed, written, engineered, and shipped by one person, on personal time and personal infrastructure, for the AGENCY 2026 hackathon in Ottawa. Special thanks to the public servants whose work makes the underlying datasets exist and accessible in the first place.

Built with
  • Next.js 16 · React 19 · TypeScript
  • ElevenLabs voice agents · Anthropic Claude
  • Leaflet + CartoDB Dark Matter tiles
  • Cloudinary for video
  • Hosted on Netlify · Source on GitHub
Public sources
  • CRA · Charities Directorate · T3010 returns
  • Treasury Board of Canada · Grants and Contributions disclosures
  • Government of Alberta · Procurement disclosures
Parallel Canada was built by Matthew Kirubakaran in a personal capacity as a private citizen. It is not affiliated with, endorsed by, or representative of any employer, government body, or political organization. The Shell Index is a research signal, not a verdict. Every score is reproducible from public sources, every weight is documented, and every flagged entity is reviewable in the demo. If a record looks wrong, please check it against the upstream source first.

Interested in building this further?

Parallel Canada is an open methodology. If you are a researcher, journalist, auditor, policy advisor, or technologist who wants to collaborate on extending this framework nationally or to new problem domains, get in touch.

Roadmap

What comes next.

Parallel Canada v1 was built in 24 hours. Here is what a fully resourced version looks like.

  1. 01
    LIVE
    Now
    Federal and Alberta datasets joined. 851,300 entities scored. Shell Index v1 operational. Four AI agents deployed. Open methodology. Cross-jurisdictional accountability infrastructure exists for the first time.
  2. 02
    IN SCOPE
    Next
    Expand to BC, Ontario, Quebec, and Manitoba. Add municipal procurement datasets. Director network analysis at scale. Shared address clustering nationally. Every province auditable within weeks of onboarding.
  3. 03
    PLANNED
    Journalism Layer
    Integration with the Investigative Journalism Foundation's 12 public databases including Revolving Door, Lobbying Communications, Government Funding, Procurement, Political Donors, Charity Tax Returns, and more. When a Shell Index flag and a lobbying communication and a political donor record all point to the same entity simultaneously, that is the story no journalist has been able to find at scale. Until now. This is the layer that connects public accountability data to public accountability journalism.
  4. 04
    FUTURE
    Scale
    Real-time data refresh as government datasets update. Public searchable registry of all flagged entities. API access for journalists, researchers, and audit offices. Whistleblower-safe submission portal for entities with insider knowledge of funding irregularities. Integration with Auditor General workflows.
  5. 05
    VISION
    Beyond
    The Parallel Canada framework is jurisdiction-agnostic and domain-agnostic. Healthcare procurement. Infrastructure spending. Municipal grants. Lobbying patterns. Foreign funding flows. Any structured dataset where public money moves and accountability is diffuse becomes auditable in hours. The methodology does not change. Only the data does.
Contact

Questions, corrections, collaboration.

Partnership, institutional deployment, media, and collaboration inquiries welcome.

Prefer email? Reach me directly at contact@parallelcanada.com