Independent project · Independent simulation
Review Desk enterprise workflow simulation
I participated in an enterprise transaction-review project at a previous employer. The original work is proprietary and confidential, so I created this independent example simulation to demonstrate the general workflow using synthetic data and invented policies.
Independent portfolio example. This is not the employer’s original system and contains no employer code, policy documents, customer data, internal forms, or integrations. My participation in the original project does not imply sole engineering ownership.
Captioned walkthrough assembled from captured interface screens. Synthetic records, invented policies, simulated AI, and a local-only Finance queue throughout.
Why I created this example
A reviewer needs to understand which records triggered a policy check, verify the source evidence, complete missing information, and send a documented case to Finance. A flag alone does not provide that complete workflow.
My contribution to the simulation
I scoped and built this independent example with AI coding assistance, defined its fictional scenario, checked the behavior in the browser, and verified the underlying rules and review state with automated tests. The example is new work; the confidential enterprise implementation is not shared or reproduced.
Tools & methods
Python · SQLite · JavaScript · CSV validation · Automated testing · AI coding assistance
These are prototype verification facts, not enterprise adoption or business-impact metrics. The AI explanation and transaction feed are clearly labeled simulations.
From incoming transaction to an explainable finding
Three fictional point-of-sale events enter the scenario. The final $6,200 cash transaction exceeds the invented $5,000 single-cash threshold. Related cash for that customer totals $8,400, exceeding the invented $7,500 threshold over seven inclusive calendar days.
The threshold checks actually execute against the synthetic records. Findings retain source transaction IDs, amounts, and the rule snapshot. A clearly labeled simulated AI explanation summarizes the evidence; no language model is called.
A prefilled form that still requires a person
The mock internal review form carries forward the customer, transaction date, receipt reference, and source amounts. The reviewer completes the business purpose, prepared-by field, review notes, and synthetic-document confirmation.
Source amounts remain locked to the transaction records. Drafts can be saved, incomplete submissions are blocked, and submitted forms cannot be silently edited. This is an internal demonstration form, not an official government form or filing.
A saved Finance task and a traceable handoff
Submitting the completed mock form creates a pending item in a local simulated Finance inbox. A snapshot of the submitted form and separate submission and Finance-trigger events are saved. No external email is sent.
A simulated Finance reviewer can approve or return the case with a note. Returned cases can be resubmitted while earlier snapshots remain retained. The case and its audit history can be exported as JSON.
How I verified the prototype
Fifteen automated tests cover rule thresholds, inclusive date windows, customer isolation, invalid records, saved review decisions, policy snapshots, incomplete forms, duplicate submissions, source-amount protection, stale edits, and Finance return/resubmission.
I also exercised CSV upload, required review notes, form completion, mock submission, reload persistence, and case export through the browser. The walkthrough is a captioned video assembled from captured screens of the working prototype.
What this example does not claim
This is a functional local prototype, not a production deployment or a claim of enterprise adoption. The feed and AI explanation are simulated. It is not the confidential employer implementation, and its fictional policies are not Form 8300 requirements.
A production version would require approved policies, authenticated identities, role-based permissions, secure integrations, protected audit history, independent security testing, monitoring, and operational ownership.
Outcome
A demonstrable workflow with clear boundaries
A working, inspectable example of a complete transaction-to-Finance review workflow, with source code, 15 automated tests, a captioned walkthrough, and explicit boundaries between functioning software and simulated components.
Download the example source and tests Download the captioned walkthroughLet’s connect
What could work better?
I’m interested in AI innovation and solutions roles where research becomes useful systems.