Skip to main content
Back to selected work

Investment Research Automation

Commercial Space & Satellite Internet Research OS

A commercial-space research system that organizes source records, conclusions, and calculated values while keeping calculated values separate from AI-written explanation.

Live static proof · Read-only public snapshot · No public model or provider connection

All flagship products
RoleAI Product Lead & Co-Developer
StatusLive static proof · Read-only public snapshot · No public model or provider connection

A commercial-space research system that organizes source records, conclusions, and calculated values while keeping calculated values separate from AI-written explanation.

Domain
Investment Research Automation
Role
AI Product Lead & Co-Developer
Current status
Live static proof · Read-only public snapshot · No public model or provider connection
Team boundary
Built in a two-person human team. The overall system is not presented as solo-built.

Problem

The user problem

Investment reviewers need to compare financial statements, operating metrics, customer and contract records, and program data without mixing currencies, losing sources, or letting AI rewrite a calculated value.

Context

Project context

Two people built and iterated the product. The investment manager and satellite-research personnel used it internally; the public site is a read-only subset of the historical product.

Workflow

Steps from question to review

  1. Define the research question and evidence scope
  2. Collect and organize financial, operating, customer, contract, and program evidence
  3. Run deterministic calculations and validation controls
  4. Keep AI narrative separate from source facts and calculated values
  5. Review sources, assumptions, conflicts, and exceptions before saving the output

Controls

Controls I chose

  • Missing values remain blank rather than being invented.
  • Conflicting records are flagged rather than silently overwritten.
  • Incompatible statements are excluded from additive totals and disclosed.
  • Human reviewers can inspect sources, assumptions, and exceptions.

Evidence

What has been verified

  • Led requirements, workflow and framework design, source collection, front-end development, release testing, deployment, and post-internship iteration.
  • Designed a cross-currency validation control that prevented an approximately 13% HKD/CNY aggregation overstatement.

Public version

Open the current public version

The link opens the public version described above. Its status and limits are the same ones listed on this page.

Continue

Compare the other two products