Accounting × AI · San José, CA

Accountant who builds financial systems alongside modern AI.

An accounting foundation paired with real engineering — building the financial and analytical tools that usually take two different people. The work runs from bookkeeping and financial analysis through Python, machine learning, and AI systems.

See the work Résumé ↗

The intersection

Accounting and software, in one skill set.

Accounting graduates rarely build software, and software engineers rarely know the books. Hiring data shows firms increasingly want both in one person, and few candidates have both. That overlap is the focus here.

The path was accounting first — to understand how money and businesses actually work — then engineering, to build the analytical and financial tools around it. The approach is practical: accounting judgment decides what's worth building, and modern AI and ML do the building.

Selected work

Things I built and shipped.

01 · Model risk Live on GitHub

Independent Model Validation — SR 11-7

I took a machine-learning trading model I built, and validated it against the bank supervisory standard for model risk (Federal Reserve / OCC SR 11-7) — two independent tools (deepchecks + giskard), a formal report, and a fully reproducible environment. The verdict was not approved for deployment, and that's the point: the real skill isn't finding an edge, it's being able to prove to a professional standard that one isn't real.

  • Python
  • LightGBM
  • deepchecks
  • giskard
  • SR 11-7
  • walk-forward validation
View repository ↗ Read the report ↗
02 · Computer vision Live on GitHub

CNN, Trained From Scratch — Object Tracking

A small CNN, trained from scratch — no pretrained weights, ~24.8K parameters, ~22ms on CPU — to track an object whose color sits below the human eye's own perceptual threshold from its background. Motion tracking and color tracking both failed for specific, measured reasons; a heatmap-regression network trained on hand-labeled frames, with a real active-learning loop, didn't. Reported honestly: strong in clear conditions, silent rather than wrong when it isn't sure.

  • PyTorch
  • Computer vision
  • Active learning
  • From scratch
View repository ↗
03 · Accounting Foundation

The accounting half

The domain the rest is built on — bookkeeping, reconciliations, journal entries, and financial statements across real small-business workflows in QuickBooks Online, plus the financial-analysis coursework behind an accounting degree at San José State.

  • QuickBooks Online
  • reconciliations
  • financial statements
  • financial analysis

The workbench

What I work with.

AAccounting & finance

  • Bookkeeping & reconciliations
  • Journal entries & the close
  • Financial statements
  • QuickBooks Online
  • Financial analysis
  • Model-risk concepts (SR 11-7)

BBuilding

  • Python
  • Machine learning (LightGBM)
  • Data pipelines & validation
  • LLM agents & orchestration
  • Retrieval / embeddings
  • Reproducible environments · Git

CThe fusion

  • AI-augmented bookkeeping
  • Independent model validation
  • Workflow automation
  • Accounting judgment → software

Résumé & portfolio

The full record.

A résumé for applications — and the work itself, documented, on GitHub.

Résumé (PDF) DOCX Portfolio ↗

Contact

Let's connect.

Based in San José — up for roles, collaborations, or just trading notes on accounting and AI.

jessealaguna@gmail.com github.com/jesselaguna ↗