Neptune Data / Bespoke Applications

Software shaped to your practice.

Every firm has one workflow that defines it — and generic tools that almost fit it. Neptune builds the application that actually does. Think of it as consulting whose deliverable is software: scoped with your practitioners like an engagement, validated against your own reference work product like a review, and delivered on a platform built for confidential work — where it keeps running long after the engagement ends.

The model

How a bespoke application gets built

It runs like an engagement: four phases, each with your practitioners in the loop. But where an engagement would end in findings, this one ends in an application whose rules encode how your firm does the work — and that's when it starts earning.

01

Scope

We sit with the people who do the work. What comes in, what goes out, what "correct" means, and which judgment calls belong to humans.

02

Build

Extraction rules, reconciliation checks and outputs modeled on your real documents — statements, schedules, side letters, the messy ones.

03

Validate

Output is scored against work product your firm has already signed off on, document by document, until the differences are explainable.

04

Operate

The app ships into your workspace with lineage, review tooling, cost visibility and your corrections feeding back into future runs.

Spotlight case study

The K-1 Extraction Suite

Built with a tax practice whose clients hold many fund interests: every spring, a mountain of Schedule K-1 packages that must become one reconciled workbook. This is what a bespoke application looks like at full depth.

The raw material

A single K-1 "package" is 40–60 pages, and most of it is noise. The suite knows the anatomy — what to read, what to cross-reference, and what to ignore entirely.

p. 3K-1 face Boxes 1–21, partner info, capital account (Item L). Primary source.
pp. 4–12Federal statements Box breakdowns, AMT detail, Item L reconciliation. Read line by line.
pp. 13–20Schedule K-3 Foreign-source income and taxes by category column. The only source for the foreign tab.

Rules, not vibes

The extraction is governed by hard rules a reviewer can hold it to — encoded from how the firm's own preparers check the work.

The totals rule

Every box's components must sum exactly to its printed total — the K-1 face amount, or the statement's own TOTAL line. A component can never stand in for a whole box, and no component gets dropped.

The column rule (K-3)

Foreign amounts are the sum of the foreign category columns — never the U.S. source column, never the grand total. The most common human error in foreign tax credit prep, eliminated by construction.

The roll-forward check

Beginning capital + contributions + income + withdrawals must reach ending capital. Anything unexplained surfaces as an explicit adjustment line — never silently absorbed.

The reconciliation self-check

Where the package includes an Item L income reconciliation, extracted line items must reproduce it before the run is accepted — even $1 items count.

What comes out

The reconciled workbook

One Excel workbook per tax year: a detail tab with one column per fund and canonical rows for every box; a foreign tax credit tab; a capital-changes partner allocation worksheet; a non-passive split. Totals are live formulas — corrections never go stale. Prior-year columns and differences appear automatically when last season's data exists.

The interactive data view

The same tabs in the browser. Every cell knows its lineage: extracted figures show the statement lines behind them and open the source PDF to the recorded page with the amount highlighted; calculated figures show their arithmetic as a process chart with every component one click away. Cycle through every place a figure appears in the package.

The entity map & analyst

A visual of strategies, funds and capital flows — phantom income and loss positions flagged — plus a conversational analyst that answers against the extracted data. Corrections agreed in review are remembered and applied to future extraction runs automatically.

How it was validated

The suite was built against real portfolios and scored against the firm's own hand-built reference workbooks — box by box, fund by fund — with every disagreement adjudicated against the source PDF until the remaining differences were explainable and documented. That validation loop never ends: reviewer corrections keep feeding the system.

Dozens
of packages per run, processed server-side while you work
Hundreds
of line items per portfolio, each tied to a printed total
Every figure
traceable to its source page — extraction records the page it read

Also built on the bespoke engine

The same discipline, other practices

Live

Loan Review

Credit agreements and loan files into structured abstracts: parties, economics, covenants, defaults and maturities — with the source language quoted.

Live

Lease Abstraction

Commercial leases into consistent abstracts: term, rent schedules, options, CAM, assignment and termination rights — portfolio-comparable.

Live

Entity & Succession Mapper

Organizational documents into an entity map with ownership, control and succession paths — the picture estate counsel draws by hand, generated.

Each of these started the same way the K-1 Suite did: a firm, a workflow, a stack of documents and a reference standard. Yours would too.

Your practice next

What's the workflow your firm is known for?

Bring the documents and the work product you're proud of. We'll scope the application together.

Start the conversation