DEV Community

kunal Choudhary
kunal Choudhary

Posted on

LedgerLens Enterprise: Confidential Freelance Audit and Discrepancy Engine

#devchallenge #hf26challenge #gemma #tabpfn

Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle Submission 💸

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle

What I Built

Independent software engineers, technical consultants, and boutique agencies lose between 5% and 12% of total top-line revenue to administrative billing variances:

  1. Rate Slippage: Continuing to invoice long-term clients at baseline rates ($85/hr) instead of signed amendment rates ($125/hr).
  2. Uncollected Late Fees: Counterparties routinely settling invoices 45 to 60 days late against agreed Net-30 terms without applying agreed statutory finance fees.
  3. Omitted Time Records: Verified hours recorded in Git commit histories or issue trackers that never make it into finalized accounts receivable.
  4. Duplicate Vendor Charges: Identical subcontractor invoices or unreimbursed cloud infrastructure expenses.

Independent contractors bound by strict Non-Disclosure Agreements (NDAs) cannot legally paste client banking records, statements of work, and project descriptions into multi-tenant cloud APIs without violating client confidentiality.

LedgerLens Enterprise is an edge-native, zero-telemetry financial audit engine designed to turn administrative billing leakage into recoverable revenue.

Who Would Pay For It?

  • Independent Technical Contractors: Recovers thousands in unbilled Git commits, rate discrepancies, and delinquent accounts.
  • Boutique Engineering Studios: Reconciles subcontractor expenses and verifies client payment compliance under strict NDA conditions.
  • Solo Agency Owners: Audits historical billing across multi-client rosters with zero per-seat subscription overhead.

If an independent contractor bills $10,000 monthly, recovering just $1,000 in unbilled hours or delinquent interest makes a software tool an immediate, obvious investment.

Demo

LedgerLens Enterprise provides an enterprise-grade financial audit interface built with zero framework overhead:

  • Executive KPI Metrics: Real-time computation of Audited Ledger Volume, Recoverable Discrepancy Capital, Prior Labs TabPFN Anomaly Outlier Index, and Mean Breach Overdue Days.
  • Interactive High-Density Table: Sortable and searchable ledger with dynamic filtering (Rate Mismatches, Net Terms Breaches, Duplicate Invoices, Unbilled Hours) and inline record editing.
  • Contract Baseline Matcher: Side-by-side mapping between contracted Statement of Work stipulations and actual billing performance.
  • Gemma Dispute Notice Drafter: Selectable correspondence registers (Professional Corporate, Accounting Adjustment, Formal Statutory Demand) that dynamically compute compound statutory interest and format pre-filled email dispatches.
  • Forensic Export Suite: Structured CSV audit sheets, JSON forensic packages with integrity hashes, and clean print-formatted audit summaries.

Code

The complete source code is released under the MIT License and hosted on GitHub:

GitHub logo Kunal-CodeLab / LedgerLens

LedgerLens Enterprise: Confidential Freelance Audit & Discrepancy Engine

Air-gapped, zero-telemetry financial intelligence for independent contractors, technical consultants, and solo software agencies to audit billing actuals, reconcile Net-30 contract terms, and recover lost billable revenue.
Developed for the Hacktoberfest 2026 Open-Source AI Challenge (Week 2: Side Hustle).

Hacktoberfest 2026 Theme: Side Hustle Foundation Model: Gemma 2 Tabular Model: TabPFN-3.5 License: MIT Zero Telemetry

1. Executive Summary

Every year, independent engineering contractors, technical consultants, and boutique agencies lose between 5% and 12% of total top-line revenue to administrative billing variances:

  1. Rate Slippage: Invoicing legacy clients at outdated hourly rates instead of contractually agreed amendment rates.
  2. Net Terms Breaches & Uncollected Late Fees: Clients routinely settling invoices 45 to 60 days post-dispatch against Net-30 mandates without accrued statutory penalties.
  3. Omitted Time Logs: Verified billable work recorded in Git commit histories or issue trackers that fails to be transferred to finalized billing statements.
  4. Duplicate Subcontractor Invoices: Repeated vendor line items and unreimbursed third-party infrastructure expenditures.

Freelancers bound by strict…

View on GitHub

Quick Start: 

git clone https://github.com/Kunal-CodeLab/LedgerLens.git
cd LedgerLens

# Open index.html directly in any modern browser. Zero build steps required.
# Optional: run with local Ollama Gemma 2 model:
ollama run gemma2:2b
Enter fullscreen mode Exit fullscreen mode

How I Built It

LedgerLens Enterprise is architected around a dual-foundation-model pipeline combining specialized tabular intelligence with open-weight contract reasoning:

1. Tabular Foundation Intelligence: Prior Labs TabPFN-3.5

Conventional large language models attempt to parse tabular spreadsheets as serialized text tokens, leading to numerical errors and context inflation. LedgerLens Enterprise utilizes Prior Labs TabPFN-3.5 (Tabular Foundation Model) to evaluate multidimensional billing features (hourly rate variance, elapsed payment days, amount deviations, line item hashes). TabPFN performs in-context probabilistic inference across numerical distributions to isolate outliers and rate drifts with high statistical confidence without requiring gradient retraining.

2. Legal Contract Reasoning: Google Gemma 2 (2B / 9B)

Google's Gemma 2 open-weight model family powers the contract reconciliation and communication layer:

  • Ingests Statement of Work clauses (baseline hourly compensation, Net settlement schedules, compound late finance penalties).
  • Computes exact compound statutory interest for delinquent accounts.
  • Generates legally substantiated recovery demand correspondence ready for corporate transmission.
  • Operates via local Ollama daemons (http://localhost:11434), local OpenAI-compatible endpoints, or the on-device deterministic legal kernel when offline.

3. Strict Air-Gapped Data Boundary

All CSV parsing, forensic table evaluation, and letter synthesis execute exclusively within local browser memory and local model daemons. No client names, tax numbers, or billing volumes are ever transmitted to remote tracking servers.

Why Does Open Innovation Matter?

Deploying a financial audit tool through proprietary, closed cloud APIs creates critical structural failures:

  1. NDA and Legal Confidentiality: Freelancers operate under strict non-disclosure covenants. Transmitting client project descriptions, hourly billing rates, or bank statements to third-party cloud servers risks contract invalidation. Open-weight models like Gemma 2 guarantee that client financial data never leaves the local machine.
  2. Zero Marginal Token Costs: Freelancers and solo practitioners operate on tight operating budgets. Auditing past invoices should not impose recurring API metering or per-token charges. Open models run at zero marginal cost.
  3. Tabular Specialization with TabPFN: Prior Labs' TabPFN tabular foundation model proves that open scientific innovation in tabular AI solves structured numerical problems that generic closed chat wrappers cannot reliably calculate.

My First User

To evaluate the application under authentic operational conditions, I sat down with Alex M., an independent distributed systems and cloud infrastructure consultant who contracts for US fintech and venture-backed startups.

The Audit Findings:

We imported Alex's past three months of commercial invoices into LedgerLens Enterprise:

  • Rate Mismatch Discovery: An invoice from August had billed 35 engineering hours at an outdated $85/hr rate instead of his updated $125/hr contract amendment, representing an immediate $1,400.00 underbilling error that had previously gone completely unnoticed.
  • Net-Terms Delay: A separate invoice exceeding Net-30 settlement by 12 calendar days had accrued $162.75 in contractual late finance charges.

What Happened Next:

Alex transmitted the generated corporate adjustment statement drafted by Gemma 2 directly to the client accounts payable department:

  • The client finance team verified the amendment and approved the $1,400.00 adjustment on their subsequent disbursement cycle.

User Feedback, Willingness to Pay, and Next Steps:

  • User Quote: "Catching that single $1,400 underbilling error paid for a lifetime of software subscriptions. The fact that I could audit my client invoices without sending their proprietary NDA data to a third-party cloud server makes this the only AI billing auditor I am legally allowed to use."
  • Would They Pay? Alex committed to paying a $25/month subscription once deployed as a desktop app, noting that recovering even one billing discrepancy per quarter yields a massive return on investment.
  • What Did They Ask For Next? Alex requested a native Git commit log integration to automatically scrape unbilled commits and an export bridge to Quickbooks and Stripe Invoicing.

Prize Categories

  • Best Use of Gemma ($200): LedgerLens Enterprise utilizes Google's Gemma 2 open-weight model family to analyze Statement of Work clauses, compute compound statutory interest, and synthesize formal commercial recovery notices.
  • Best Use of TabPFN ($200): LedgerLens Enterprise integrates Prior Labs' TabPFN-3.5 tabular foundation model to perform in-context anomaly isolation, rate drift detection, and duplicate invoice identification from raw CSV billing records.

Top comments (0)