Jeremy Helvey, CPA

AI Finance Analytics Warehouse

Analytics engineering + an LLM natural-language query layer

An end-to-end analytics warehouse that transforms raw general-ledger, project, and chart-of-accounts data into decision-ready models — an income statement, project profitability, and budget-vs-actual — using dbt on DuckDB. On top of it sits a natural-language layer: ask a plain-English question and an LLM writes validated, read-only SQL, runs it, and returns the answer.

• Accounting-aware data tests — every journal entry must balance (debits = credits) before a model is allowed to build.
• Guardrailed AI — the model only ever sees the schema, can only issue read-only SELECTs, and every generated query is validated before it runs.
• Reproducible & documented — version-controlled transformations with auto-generated lineage and column-level documentation.

Small-Business Accounting & Reporting Platform

A standalone double-entry accounting application, built in Go

An all-in-one accounting and reporting tool for a small practice. It imports and parses revenue and expense data (including QuickBooks exports), reconciles everything into a double-entry general ledger that re-posts from source on every change, and produces P&L, chart-of-accounts, cash-flow, and monthly reports with PDF and Excel export.

• Allocation engine — splits shared overhead across staff by configurable rules (all, by state, by employment type), with per-expense overrides.
• Data hygiene — fuzzy name-matching and alias consolidation, duplicate detection, and a corrections workflow that re-posts the ledger.
• Ships as a single binary — launches a local web app in the browser; no install, no cloud dependency.

Runs entirely on the local machine. Real client data never leaves the device — the screenshot above and any public demo use a fully synthetic dataset.