You will get a multi-language product matching pipeline with AI fallback
Rising Talent
Project details
Most matching systems rely on exact EAN lookups (missing 40%+ of titles) or pure AI (which hallucinates). This pipeline combines both with zero false positive auto-accepts.
It solves a real problem: mapping variable supplier titles to 3,000+ products across 7 suppliers and 4 languages. The same product appears as "SAMSUNG A15 128GB BLK 4G", "Galaxy A15 4G 128GB Zwart" (Dutch), and "SM-A155 Galaxy A 128GB Schwarz" (German). Each must match correctly or route to human review.
4-stage pipeline: EAN exact lookup, rule-based title parsing (55 brands, 151 families, 97 color aliases across 4 languages), attribute scoring with hard gates (brand or storage mismatch = instant disqualify), and AI fallback for garbled titles. AI matches always go to human review. Operator corrections become reusable rules automatically.
237 tests, Docker deployment, full API documentation, production-ready with MSSQL integration.
It solves a real problem: mapping variable supplier titles to 3,000+ products across 7 suppliers and 4 languages. The same product appears as "SAMSUNG A15 128GB BLK 4G", "Galaxy A15 4G 128GB Zwart" (Dutch), and "SM-A155 Galaxy A 128GB Schwarz" (German). Each must match correctly or route to human review.
4-stage pipeline: EAN exact lookup, rule-based title parsing (55 brands, 151 families, 97 color aliases across 4 languages), attribute scoring with hard gates (brand or storage mismatch = instant disqualify), and AI fallback for garbled titles. AI matches always go to human review. Operator corrections become reusable rules automatically.
237 tests, Docker deployment, full API documentation, production-ready with MSSQL integration.
Programming Languages
PythonCoding Expertise
Performance Optimization, SecurityWhat's included
| Service Tiers |
Starter
$500
|
Standard
$2,000
|
Advanced
$4,000
|
|---|---|---|---|
| Delivery Time | 3 days | 5 days | 7 days |
Number of Revisions | 0 | 0 | 0 |
Design Customization | - | - | - |
Content Upload | - | - | - |
Responsive Design | - | - | - |
Source Code |
Optional add-ons
You can add these on the next page.
Additional Revision
+$45Frequently asked questions
About Gozde
AI Systems Architect | Mechanical Engineer + Process Automation
Istanbul, Turkey - 8:16 pm local time
Mechanical Engineer turned AI Systems Architect. 8 years across mechanical design, manufacturing and quality engineering, and technical sales in medical devices, packaging automation, and precision manufacturing. Now building intelligent automation systems that bridge engineering, AI, and business operations.
Recent project highlights:
Multi-agent e-commerce system - 4 autonomous Python agents handling Amazon, Walmart, Shopify, and 35+ vendor portals. Full order lifecycle from intake to tracking confirmation.
Product matching pipeline - 4-stage system mapping supplier titles to 3,000+ products across 4 languages. 237 tests, zero false positive auto-accepts.
AI documentation agent - generates user guides from Linear issues, changelogs from releases, and QA reports. RAG-enhanced with human-in-the-loop approval.
Generative design pipeline - physics simulation, surrogate modeling, and evolutionary optimization producing validated CAD output.
Digital transformation of a sheet metal manufacturing SME - from zero infrastructure to automated inventory, order management, and reporting. Eliminated 8-12 hours per week of manual coordination at zero monthly cost.
Tech: Python, FastAPI, Claude API, n8n, Playwright, Docker, Airtable, Voiceflow, SolidWorks, ANSYS.
I work remotely, manage projects independently, and deliver systems meant to be handed off and operated without me.
Steps for completing your project
After purchasing the project, send requirements so Gozde can start the project.
Delivery time starts when Gozde receives requirements from you.
Gozde works on your project following the steps below.
Revisions may occur after the delivery date.
Step 1: Data analysis
- review your product catalog structure, supplier file formats, and matching requirements
Step 2:
Configure catalog loader and supplier ingestion for your specific data schema