For operations, eCommerce, catalog, and the data workflows in between.
Turning operational complexity into structured, automated systems.
Building structured, automated systems out of messy data, scattered catalogs, and one-off operational workflows. Quietly, without overengineering. The background is product data, marketplace onboarding, technical implementation, SOPs, dashboards, and the kind of operational troubleshooting that happens after a tool was supposed to "just work." That work is now paired with practical LLM tooling: Claude Code, ChatGPT, Gemini, and Google Antigravity used to ship AI-assisted Python, Apps Script, SQL, and workflow automations. Working systems instead of slide decks. The work tends to land where commercial, operations, data, and engineering teams stop talking to each other.
E-Commerce & Management:
Shopify, WooCommerce, Mirakl, Rithum/DSCO, Radial, LogicBroker, Cymbio App, Airtable, Asana, Monday.com.
Data & Automation:
Claude, Claude Code, ChatGPT, Gemini, Google Antigravity, AI-assisted Python automation, Advanced Excel, SQL, Power BI, Google Apps Script, Make.com, Zapier, Postman, KPI dashboards.
Problem: a live product catalog needed product, variant, model, image, taxonomy, material, color, size, and multilingual metadata pulled into one PIM-ready structure.
Solution: a Python multi-agent scraper and export workflow, adaptable to website and marketplace catalogs, with discovery, enrichment, checkpoint/resume logic, dashboard monitoring, and structured CSV outputs.
Impact: 3,446 product URLs processed into 39,776 variant-level rows, ready for review, enrichment, and export.
View pipeline UI PreviewProblem: high-volume review and cleanup work was bottlenecking a single linear pipeline.
Solution: a parallel AI-assisted workflow splitting the work across specialized agents with structured outputs and review logic.
Impact: ~80% reduction in processing time, with the work still organized enough to audit and improve.
Problem: manual categorization of new products was slow, inconsistent, and didn't scale across catalogs.
Solution: an AI-assisted classifier that scans titles and descriptions, then maps products to categories, attributes, and taxonomy logic.
Impact: ~80% reduction in manual categorization work, plus a reusable pattern for catalog cleanup that carries to other projects.
Problem: a boutique brand with ~18,500 product variants was managing a complex catalog through disconnected spreadsheets and manual review.
Solution: an interactive AI-first PIM prototype showing how product data, images, categories, attributes, and enrichment workflows could become one structured operating system.
Impact: a broad operational problem converted into a concrete product concept, giving the brand a practical direction for catalog cleanup and workflow automation.
Problem: Project Banana is a long-term PIM vision, but phase one needed to stand on its own: extract and organize product data from a live catalog.
Solution: an AI-assisted Product Data Engine combining spreadsheet data with live website scrapes, then structuring SKUs, models, images, categories, descriptions, colors, and attributes into PIM-ready outputs.
Impact: the first working module of Project Banana, plus a standalone product-data tool for catalog extraction, enrichment, and PIM readiness.
Problem: a homeware brand was expanding from a single WooCommerce catalog onto 4 UK marketplaces running on Mirakl platform (B&Q, Debenhams, Tesco, The Range). Each retailer required different templates, taxonomy, image rules, and SEO standards.
Solution: an AI-assisted catalog automation workflow that mapped one WooCommerce catalog through Mirakl into 4 retailer-ready templates, each shaped differently. Handled product normalization, category mapping, compliance fields, image structuring, retailer-specific attributes, and SEO-optimized content.
Impact: 4 marketplace-ready templates delivered, 441 fields mapped, 2,414 SEO text items rewritten, and average SEO scores up from ~57 to ~90 out of 100.
Problem: needed a lightweight pipeline-tracking tool that wasn't a heavy Salesforce-style setup, just enough to keep status, notes, and next actions in one place.
Solution: a practical Mini CRM built with AI-assisted development, lead status tracking, notes, next actions, and a simple follow-up structure.
Impact: a focused internal tool that keeps pipeline data organized, and a working demonstration of turning a personal ops problem into a usable workflow application.
Open live demoProblem: a standard CV doesn't show the practical work behind AI automation, data operations, and internal-tool building.
Solution: this portfolio website itself, deployed as a live proof-of-work layer and built with AI-assisted development to structure the story, the pages, and a recruiter-friendly project format.
Impact: a working example of shipping functional websites end-to-end, translating complex operations work into clear product-style pages, and using web delivery as part of a personal-brand system.
Small projects with clear scope and a defined output. Each one solves a specific data, catalog, or operations problem and starts in days, not quarters.
One messy Excel, CSV, or Google Sheets mapping file: cleaned, validated, and ready for upload, review, migration, or a larger workflow.
Ask about Excel mappingMessy lists, exports, CSVs, or spreadsheets: deduplicated, formatted, and structured so the file becomes easier to use and easier to hand off.
Ask about data cleanupBroken or missing formulas, lookups, validation rules, calculated columns, and spreadsheet logic: repaired or rebuilt inside the existing file.
Ask about Excel formulasA confusing tracker or recurring report: rebuilt into a cleaner spreadsheet with clearer tabs, fields, summaries, and next actions.
Ask about report cleanupProduct data extracted from a live ecommerce catalog and delivered as a clean CSV/XLSX, ready for migration, PIM, Shopify, WooCommerce, or marketplace work.
Ask about catalog rescueProduct information pulled from live scraping, customer files, supplier sheets, and image links, then mapped into a ready-to-go catalog or retailer upload template.
Ask about product migrationMessy product data mapped into retailer or marketplace templates: required fields, taxonomy, attributes, images, and content cleanup, all in one pass.
Ask about marketplace templatesOne repeated manual workflow becomes a first usable AI-assisted version: script, dashboard, SOP, or internal tool.
Ask about workflow automationAI-assisted workflows, product data systems, and automation pipelines for commerce, marketplace, catalog, and back-office teams.