RISK INTELLIGENCE / PROJECT 05

AV / CITY

Interrogate the evidence behind autonomy.

Explore a historical California autonomous-vehicle collision snapshot through geographic filters, source references, and explicit statistical qualifications.

DEVELOPMENT STATUSWorking historical dashboard

APPLICATIONAutonomous mobility · risk intelligence

TECHNOLOGY

Interactive HTMLEmbedded dataSVG mapsStatistical context

Explore the demo

VIEWING DEPTHOverview Mode (Executive Summary)

30-SECOND EXECUTIVE BRIEF

KEY TAKEAWAYS AT A GLANCE

WHAT IT DOES

Presents 259 historical California autonomous-vehicle collision records (2014–2020) across 4 analytical workspaces, 9 interactive filter dimensions, SVG geographic maps, and record-level source locators.

PROBLEM IT SOLVES

Public autonomous-vehicle crash tallies are frequently misused to rank operators or claim causal risk without checking mileage-denominator compatibility, geocoding boundaries, or reporting bias.

WHY IT MATTERS

Demonstrates rigorous risk-intelligence discipline: pairing exploratory data visualization with explicit statistical guardrails (Fisher exact tests, Wilson intervals, Holm adjustment, and denominator audits).

MY CONTRIBUTION & ROLE

Analytical framing, exposure-denominator audit logic, geographic and injury-field qualification structure, and standalone zero-backend dashboard packaging.

01 / OVERVIEW

The idea in practice.

A self-contained analytical dashboard that brings a historical collision-record snapshot, geographic exploration, descriptive charts, and evidence qualifications into one browser experience.

02 / THE PROBLEM

A real systems question.

Collision counts are easy to overinterpret. Without compatible exposure data, consistent injury fields, and a clear account of geographic uncertainty, comparisons can imply safety conclusions the records cannot support.

03 / MY APPROACH

Make the reasoning inspectable.

Embed the snapshot and interface together, expose filtering and record-level source locators, and keep limitations alongside the results. Treat data interpretation as part of the product experience.

Portfolio contribution context: domain-led concept development and AI-assisted prototyping, as described by the owner. Individual implementation authorship is not documented in the supplied files.

04 / SYSTEM ARCHITECTURE

From inputs to insight.

Select any stage below to inspect its inputs, state transformations, outputs, and failure/uncertainty handling:

STAGE 01 DATA-FLOW SPECIFICATION // HISTORICAL SNAPSHOTAutonomous Vehicle Crash Data/av-city-dashboard-upload/index.html

Inputs & State

259 reviewed collision records (14 Oct 2014 – 15 Jun 2020) with documentary references and workbook row/range locators

Processing & Logic

Separates 241 map-eligible records (including 148 in San Francisco) from 18 geographically unmatched records

Outputs & Interfaces

Self-contained chunked JSON data payload embedded directly inside the standalone 4.4 MB HTML export

Assurance & Failure Handling

Preserves the 18 unmatched records in tabular counts and audits rather than silently dropping or fabricating coordinates

TECHNOLOGY STACK

Interactive HTMLEmbedded dataSVG mapsStatistical context

Algorithms & Engineering Logic

Conditional Contingency Analysis & Holm Multiple-Testing Adjustment

Autonomous Vehicle Crash Data/av-city-dashboard-upload/index.html · docs/RIPPLE-AV-AUDIT.md

P(Characteristic | Reported Collision) with Wilson 95% CI, Fisher exact p-value, and Holm step-down correction

Evaluates descriptive associations within the reported collision set (e.g., collision counterpart or movement state conditional on a report being filed) while controlling family-wise type-I error across multiple exploratory comparisons.

Exposure-Denominator Compatibility Gate

Autonomous Vehicle Crash Data/av-city-dashboard-upload/index.html

Rate = Collisions / Miles is UNDEFINED unless Geographic_Scope(Collisions) ≡ Geographic_Scope(Miles) ∧ Mode_Definition matches

Audits why combining California DMV collision reports with annual disengagement mileage summaries without matching municipal boundaries, time windows, and conventional-mode exclusions produces invalid safety rankings.

Architectural Decisions & Engineering Trade-Offs

Single-file 4.4 MB embedded HTML/JS/data export vs. cloud database API

ENGINEERING RATIONALE

Guarantees permanent, zero-latency browser execution on any static host (including GoDaddy cPanel) with zero server maintenance or database credentials.

EXPLICIT TRADE-OFF & BOUNDARY

Larger initial iframe download (loaded lazily only when the visitor clicks Launch Demo) and a fixed historical snapshot (2014–2020).

Foregrounding methodological limitations alongside executive charts

ENGINEERING RATIONALE

Demonstrates genuine risk-assurance judgment: knowing what a dataset cannot prove is as important as visualizing what it contains.

EXPLICIT TRADE-OFF & BOUNDARY

Avoids sensational "safest operator" headlines in favor of defensible, audit-ready analysis.

Verification, Testing & Evidence Scope

EXECUTED TEST & AUDIT SUITEByte-for-byte SHA-256 integrity check of supplied 4,447,007-byte standalone export + structural parser & secret scan

259 historical records, 241 map-eligible records, 148 San Francisco records, 9 filter dimensions, 4 workspace tabs, and outbound reference domain allowlist.

WHAT THIS VERIFICATION PROVES

Self-contained browser execution, zero embedded credentials or local file leaks, and faithful preservation of all statistical qualifications.

WHAT REMAINS UNPROVEN / OUT OF SCOPE

Does not independently recalculate the upstream workbook scripts (not supplied in folder) or provide post-2020 live collision feeds.

WANT TO INSPECT THE ALGORITHMS, FORMULAS & ARCHITECTURAL TRADE-OFFS?Switch to Technical Deep Dive Mode to view 2 mathematical/logic specifications, trade-offs, and test verification details.

HANDS-ON EXPLORATION

Explore the interactive lab & working prototype.

Open full workspace in new tab

Historical collision snapshot covering 2014–2020. Reported-collision associations are not crash-risk estimates or causal effects. Exposure-denominator compatibility is unproven; no live refresh or AI service is provided.

STATISTICAL INTEGRITY & PROVENANCE // 259 HISTORICAL RECORDS

Evidence Qualification & Exposure-Denominator Audit Explorer

241 MAP-ELIGIBLE / 18 UNMATCHED

Explore how the dashboard structures statistical evidence before launching the full 4.4 MB standalone analytical workspace below. Select an analytical lens to see how geographic eligibility, injury-field reconciliation, and exposure-denominator limits are enforced.

SELECTED SUBSET259 records14 Oct 2014 – 15 Jun 2020

MAP-ELIGIBLE241 recordsVerified coordinate assignment

UNMATCHED / AUDIT ONLY18 recordsPreserved in tabular counts

FILTER DIMENSIONS9 dimensionsYear, operator, mode, injury, etc.

COMPLETE REVIEWED SNAPSHOT (2014–2020)

What the Records Support

All 259 collision reports filed under California DMV autonomous vehicle reporting rules between 14 Oct 2014 and 15 Jun 2020. 18 records lack street-level coordinates suitable for municipal boundary assignment.

ASSURANCE QUALIFICATION

Why Rate Rankings Are Rejected

Raw report totals reflect fleet size, urban operating hours, and reporting rules—not per-mile crash risk.

Runs the compiled application locally in your browser · no credentials or backend needed

05 / RESULTS & CAPABILITIES

What the project does.

259 embedded historical collision records, with geography eligibility clearly distinguished.
Four tabs, nine filtering dimensions, and record-level inspection.
Interactive maps, descriptive comparisons, and source references.
Visible qualifications for injury-field conflicts, geography, and exposure denominators.

06 / ENGINEERING CHALLENGES

Where the hard questions are.

Preventing record counts from being interpreted as collision rates.
Exposing inconsistent fields and approximate geography.
Keeping historical data freshness distinct from the dashboard export date.

Current boundaries

The snapshot covers 14 October 2014 through 15 June 2020; the export date does not refresh the records.
Statistical qualifications are documented, but this portfolio does not independently reproduce every statistic; the underlying workbooks and analysis scripts were not supplied.
Associations are conditional on being a reported collision and cannot establish crash risk, causation, or operator safety rankings.
AI handoff features belong to the original host environment and are unavailable in standalone hosting.

07 / FUTURE DEVELOPMENT

The next useful steps.

01Add a reproducible refresh and provenance pipeline.
02Validate exposure compatibility before introducing rate-based comparisons.
03Independently reproduce statistical outputs from the underlying snapshot.

08 / AI DEVELOPMENT TRANSPARENCY

Tools, judgment, and evidence.

The portfolio owner uses ChatGPT, Claude, and Google Antigravity for concept development, design, and prototyping. Project-specific records do not establish which tool or person authored individual components. No precise human/AI coding split is claimed.

The capabilities described here were checked against supplied source files. Model behavior, cached examples, historical data, and proposed features are identified separately.

Source evidence

Autonomous Vehicle Crash Data/av-city-dashboard-upload/index.html
Autonomous Vehicle Crash Data/av-city-dashboard-upload/UPLOAD-INSTRUCTIONS.txt

Local source references for provenance; confidential materials are not distributed.

NEXT IN THE COLLECTION

SYSTEMS STUDY

Connecting the whole industrial system.

THE NEXT CONVERSATION

Let’s work on
what comes next.

Technical programs. Systems assurance. Intelligent technology.
Complex problems with a practical purpose.

Connect on LinkedIn

Interrogate the evidence behind autonomy.

The idea in practice.

A real systems question.

Make the reasoning inspectable.

From inputs to insight.

Historical snapshot

Filter state

Local visualizations

Interpretation layer

Algorithms, Trade-Offs & Verification Evidence

Algorithms & Engineering Logic

Conditional Contingency Analysis & Holm Multiple-Testing Adjustment

Exposure-Denominator Compatibility Gate

Architectural Decisions & Engineering Trade-Offs

Single-file 4.4 MB embedded HTML/JS/data export vs. cloud database API

Foregrounding methodological limitations alongside executive charts

Verification, Testing & Evidence Scope

Explore the interactive lab & working prototype.

Evidence Qualification & Exposure-Denominator Audit Explorer

What the Records Support

Why Rate Rankings Are Rejected

What the project does.

Where the hard questions are.

Current boundaries

The next useful steps.

Tools, judgment, and evidence.

SYSTEMS STUDY

Let’s work on
what comes next.

Interrogate the evidence behind autonomy.

The idea in practice.

A real systems question.

Make the reasoning inspectable.

From inputs to insight.

Historical snapshot

Filter state

Local visualizations

Interpretation layer

Algorithms, Trade-Offs & Verification Evidence

Algorithms & Engineering Logic

Conditional Contingency Analysis & Holm Multiple-Testing Adjustment

Exposure-Denominator Compatibility Gate

Architectural Decisions & Engineering Trade-Offs

Single-file 4.4 MB embedded HTML/JS/data export vs. cloud database API

Foregrounding methodological limitations alongside executive charts

Verification, Testing & Evidence Scope

Explore the interactive lab & working prototype.

Evidence Qualification & Exposure-Denominator Audit Explorer

What the Records Support

Why Rate Rankings Are Rejected

What the project does.

Where the hard questions are.

Current boundaries

The next useful steps.

Tools, judgment, and evidence.

SYSTEMS STUDY

Let’s work onwhat comes next.

Let’s work on
what comes next.