AI Product Manager & Builder · Accounting-trained

I build AI products for investment research, consulting delivery, and spreadsheet review.

Each product shows where the source came from, how a number was calculated, which exception was raised, and who makes the final decision.

View the three systems View web résumé

How I design them

Keep the source. Show the calculation. Flag the exception. Let a person decide.

Selected products

Three products, explained by the decision they support.

Each case states what the user needed, what I built, the control that mattered, the available evidence, and my exact role.

Investment Research

Commercial Space & Satellite Internet Research OS

Evidence → decision support

User decision

Help investment reviewers compare financial, operating, customer, contract, and program evidence without losing source lineage or masking incompatible data.

What I built: A cross-market research system that puts source records, calculated results, assumptions, and exceptions in one review path.
Control that mattered: Keep sourced facts and deterministic calculations separate from AI-authored narrative; leave missing values blank and surface conflicts or exclusions.
Verified evidence: A cross-currency guard I designed excluded incompatible statements and prevented an approximately 13% HKD/CNY aggregation overstatement.

My role

Led requirements and framework design, source collection, front-end development, testing, deployment, and post-internship iteration in a two-person human team.

What is live

Internal-use product with a static, read-only public proof. No investment action, return, or quantified adoption is claimed.

Consulting Delivery

AI Industry Research OS

Source → checked report

User decision

Turn a recurring AI-industry brief into a delivery where reviewers can trace report claims to source records, challenge evidence, and review decisions.

What I built: A delivery workflow for source audit, fact and viewpoint checks, challenge evidence, two review rounds, and fixed reports.
Control that mattered: Require validation and challenge evidence before a report is finalized.
Verified evidence: The first-month delivery connected a 262-source audit, an 88-item evidence-validation table, five thematic reports, and 57 of 57 internally closed second-round content checks.

My role

Translated research pain points into scope and testable requirements, co-developed the workflow, coordinated researchers and engineers through two iterations, and supported QA and handoff.

What is live

Joint iResearch/CSDN delivery. The linked LaurenceYang site is a separate sanitized, read-only snapshot. It is not a live service or evidence of client acceptance or adoption.

Financial Controls

Excel Intelligence & Controls

Workbook → review map

User decision

Help finance reviewers understand a formula-driven workbook’s structure, formulas, and exceptions before relying on its outputs.

What I built: A browser-local Analyzer that profiles structure, groups relative formula families, preserves Excel errors, maps dependency evidence, and exports a filterable review map.
Control that mattered: Do not turn a static signal into an accounting conclusion: preserve source facts, downgrade unresolved references, and require a person to approve definitions or actions.
Verified evidence: The live Analyzer keeps workbook processing on-device. The stable v1.34.0 local authority has five Skills and 76 contracts; 39 product-core checks and the 23-command synthetic smoke passed.

My role

Owned product scope, workflow, control model, public boundary, evidence design, and release QA as the independent creator/owner.

What is live

Working synthetic demonstration. No production adoption, automated posting or approval, savings, or business impact is claimed.

Background

The finance and product background behind the work.

Accounting

Bachelor of Science in Accounting

University of Florida · Fisher School of Accounting · Expected December 2027 · GPA 3.72 · AI certificate in progress

Experience

Research, delivery, and product ownership

Hands-on investment research, business–technology coordination on an AI-industry consulting delivery, and independent financial-control product development.

How the work was built

Different team shapes, explicit boundaries

Two-person product collaboration, shared research and engineering delivery, and independent product ownership are identified separately in every case.

The web résumé puts dates, roles, team boundaries, verified work, and limitations in one place.

View web résumé