h
hudsonliu007

hudson liu

@hudsonliu007

Data Automation Engineer

China
English, Chinese
About me
I build batch pipelines that clean, classify, and enrich product and ERP data. I specialize in mapping SKU titles to marketplace taxonomies, normalizing material master records in ERPNext, and running automated classification jobs with confidence scoring to optimize manual review processes.... Read more

Skills

h
hudsonliu007
hudson liu
Offline • 

See my services

Product Upload
I will categorize 1000 product skus into your taxonomy

Portfolio

Work experience

CSCEC_Egypt

CSCEC Egypt

Full-time • 3 yrs 6 mos

Data & Automation Consultant

Jan 2025 - Oct 2026 • 1 yr 9 mos

Positioning Tonglian operates in an industry where the constraint is rarely production capacity. It is trust, fragmented data and fragmented demand: every mixing plant keeps its own material codes, every region prices differently, and every customer relationship lives in someone's phone. My mandate was to turn that fragmentation into an operating advantage — a brand the market recognises, one set of numbers the group can act on, AI that removes repetitive judgement without removing accountability, and strategic accounts that buy outcomes instead of tonnage. ──────────────────────────────────── Group digital middle platform Situation. Growth by acquisition and regional expansion left the group with 30 ERP instances and 5 definitions of the same material. Month-end consolidation ran on exported spreadsheets; the same customer existed under four codes; nobody could answer "what is our margin by product line and region" without a week of manual work. What I did. I treated the platform as a governance programme with software attached, not an IT project: • Master data first. Defined one code per material, customer, plant and contract; appointed a data owner per domain; published a dictionary with definitions and change control. Nothing was migrated before the definitions were signed off. • Ingestion layer. Built connectors from ERPNext/Frappe and legacy systems into a governed store, with validation rules that rejected rather than silently accepted bad rows.

Data Pipeline Engineer — Domain Corpus & Structured Datasets

Jan 2025 - Oct 2026 • 1 yr 9 mos

Built the data pipeline behind a domain-specific engineering model: collected and licensed technical literature and standards, extracted tables and specification figures from PDFs, de-duplicated and quality-filtered the corpus, and produced both a training dataset and a held-out evaluation set. Kept deterministic numeric rules outside the model and served them from external tools. Delivered a traceable source list and an exclusion log for every rejected document.