Back
ProjectCase Study

Aperture Financial Intelligence

Evidence-led financial research with human approval

A full-stack financial research workspace that turns private source documents into cited, reviewable analysis without crossing into trade execution or personal advice.

Year2026
ExperienceLive guided demo
CategoryFintech systems + trustworthy AI
Next.jsGoSupabaseAI evaluation
Open guided demo
Aperture Financial Intelligence interface preview
ContextThe work

Research with a visible chain of evidence.

Aperture keeps source documents, retrieval evidence, structured analysis, human edits, and reviewed export in one workflow. The product does not try to make uncertainty disappear. It makes the work easier to challenge.

01

The problem

Turn private financial source material into structured research that remains traceable, challengeable, and human-approved without implying trade execution or personal advice.

02

The user

Research-minded investors and analysts who need evidence, freshness, uncertainty, and decision boundaries visible in one workflow.

03

My role

Independent product direction, UX/UI, Next.js frontend, Go API integration, AI evaluation, and release engineering

ProofDecisions

The choices that made the product hold together.

01

Separated Next.js product rendering

Separated Next.js product rendering from the authoritative Go finance and document API.

02

Bound material claims to

Bound material claims to the active retrieval evidence or a visible unsupported state.

03

Stored immutable raw output

Stored immutable raw output and append-only human edits, decisions, and citation feedback.

04

Kept external providers disabled

Kept external providers disabled by default while deterministic evaluation and failure handling matured.

ScreensSelected work

The interface, in context.

Aperture Financial Intelligence research workspace
The research workspace keeps citations and review state close to the analysis.
ReleaseProof and boundaries

What the build makes visible.

Verification

  • 40 of 40 evidence regression cases pass across five source documents.
  • The disposable-database E2E flow passes from upload through reviewed export.
  • Cross-owner isolation, prompt injection, schema repair, accessibility, and deployment contracts are automated release gates.

Limitations

  • The live guided demo is read-only and uses fictional evidence; authenticated document ingestion and reviewed export remain a private evaluation workspace.
  • External model calls and paid-provider comparisons remain disabled pending explicit cost and credential approval.
MoreCurrent case studies

Explore the other current projects where product thinking is carried through interface, architecture, testing, and release.