Business Operations
Real-world experience in billing, finance workflows, documents and process logic.
I build AI systems that turn repetitive work into automated workflows.
Billing Analyst @ Michelin. AI automation builder focused on documents, data, finance operations and practical workflow intelligence.
Python Automation
AI Agents
Document Intelligence
[ 01 — WHO I AM ]
Automation is where I think best.
I work at the intersection of business operations, finance processes and artificial intelligence. My daily world is structured data, documents, repetitive workflows, Excel logic and operational precision. My passion is turning that world into automated systems that save time, reduce manual work and make processes easier to control.
> currently building: document pipelines · matching engines · AI agents
AI Mindset
I use AI not only for answers, but for architecture, analysis, testing, refactoring and building smarter systems.
Automation Builder
I focus on practical tools: converters, validators, matching engines, dashboards and AI-assisted workflows.
[ 02 — LIVE DEMOS ]
Working automation demos, built on synthetic data.
Public demos showing how messy operational input can become structured, validated and repeatable output. No company data, no confidential workflows — only safe examples of real automation patterns.
document intelligence
Invoice PDF → Excel Validator
A browser-based document automation demo that parses sample invoice text into structured output and runs validation checks before manual review.
sample_invoice_0248.pdf
live JS engine
PDF / extracted fields
Invoice NoINV-0248
Date2026-07-06
CustomerDemo Customer Alpha
Net€1,240.00
VAT€235.60
Total€1,475.60
extractnormalizevalidateexport
- VAT calculation: OK
- Total match: OK
- Duplicate invoice: No duplicate
- Missing PO number: Review
| Invoice | Net | VAT | Status |
|---|---|---|---|
| INV-0248 | €1,240 | €235.60 | Review |
clean XLSX generated + 1 review warning
credit control workflow
AR Dunning & Cash Collection Assistant
A browser-based accounts receivable demo for credit control workflows. It prioritizes overdue accounts, assigns dunning stages and generates reviewable follow-up drafts.
receivables_sample.csv
live JS engine
Northwind Demo€4,25012dMedium
Atlas Sample Ltd€9,80037dHigh
Blue River Test€1,1205dLow
Nova Retail Demo€6,43024dMedium
1-7d: Gentle reminder
8-30d: Dunning follow-up
31+d: Escalation
Total overdue€21,600
Highest riskAtlas Sample Ltd
Next actionescalation email
draft / not sent
Subject: Payment follow-up for invoice AR-1002
Hello, this is a polite follow-up regarding invoice AR-1002. Please review the outstanding balance and confirm the expected payment date.
process intelligence
Workflow Bottleneck Analyzer
A browser-based process intelligence demo that analyzes synthetic workflow metrics, detects bottlenecks and suggests automation opportunities.
workflow_metrics_sample.log
live JS engine
Intake1.2h
Approval9.4h
Validation2.1h
Correction6.8h
Export0.7h
Longest stepApproval, 9.4h
Main delayWaiting for approval, 46%
Rework driverMissing data, 31%
Add pre-validation before approval.
Auto-route complete requests.
Create an exception queue for missing data.
[ 03 — SELECTED WORK ]
Automation areas I explore
These are the broader automation areas behind my working demos and personal systems.
Document Automation
PDF-to-Excel converters, structured extraction, document cleanup and standardized outputs for operational workflows.
in production · used weekly
Invoice & Data Matching
Deterministic matching logic for business documents, price data, quantities, references and validation rules.
deterministic · zero-guess
AI Process Analysis
Using AI to map workflows, detect weak points, improve procedures and design better automation systems.
ongoing practice
Diet Planning Engine
A personal AI-assisted nutrition system with meal planning logic, macro adjustment, verification, testing and reproducible outputs.
personal system · daily use
Historical & Market Regime Research
Research tools for comparing market conditions, historical analogs, stress signals and regime shifts.
research mode
AI Idea Pipeline
A system for tracking GitHub trends, emerging repositories, papers and new project opportunities.
scanning daily
[ 04 — PHILOSOPHY ]
Not just prompts. Systems.
I believe the next advantage will not come from using AI casually. It will come from people who understand processes deeply enough to rebuild them with AI, automation, validation and clear architecture.
[ 05 — STACK ]
My automation stack
Everything orbits one goal: turn messy input into validated output.
AI WORKFLOW
Engine
Python
AI agents
PDF parsing
Data validation
FastAPI
Excel automation
Prompt engineering
DuckDB
CSV workflows
Power Automate
Process mapping
Testing logic
Workflow architecture
[ 06 — EXPERIENCE ]
Business context meets AI execution.
Working in a large international company taught me one thing: real automation is not about flashy demos. It is about understanding messy processes, edge cases, documents, approvals, exceptions and people. That is where useful AI systems are built.
Billing Analyst @ Michelin
Large-scale finance operations: billing, documents, deadlines and operational precision.
Finance & document workflows
Invoices, references, validation logic and the unglamorous reality of reporting.
Automation projects
Converters, matching engines and checkers — tools that quietly remove manual steps.
AI-assisted process design
Mapping how work actually happens, then rebuilding it with AI in the loop.
Personal AI tools & research systems
Diet engine, market research, idea pipeline — a lab that never really closes.
[ 07 — CONTACT ]
Let’s automate
the boring parts.
I am interested in AI automation, workflow intelligence, business process optimization and practical tools that turn repetitive work into systems.
direct line → krapcys1@gmail.com