OCR + LLM ｜ Quick PoC to validate accuracy using your actual documents
Contracts, inspection reports, maintenance logs, and other PDF documents can be safely digitized and structured with AI.
I help companies validate how accurate AI can be—before they commit to a full-scale implementation.
15+ years of engineering experience across Australia and Japan.
From evaluation to improvement and deployment, I support the entire workflow end-to-end.
👉 Book a quick consultation (15 min)
What I Offer
3–5 Day Quick PoC (Proof of Concept)
Validate whether your documents can be automated with:
- AI-powered extraction (OCR + LLM)
- Accuracy evaluation (numeric scoring)
- Error analysis & improvement suggestions
- Structured output (JSON / CSV / DB-ready)
You’ll receive a fully working demo environment based on your actual documents.
Why This Matters
Many companies face:
- Low OCR accuracy
- High manual workload
- Unclear if AI automation is feasible
- No metrics to evaluate model quality
- “We don’t know where to start”
My PoC clarifies what can be automated and how accurate it will be, so you can make a confident decision.
How It Works
Document ingestion
↓
AI extraction
↓
Accuracy evaluation
↓
Improvement
↓
Structured data output
You’ll see exactly which fields are accurate, which ones need tuning, and how much improvement is possible.
About Me
Norihiro Ueki
Sydney-based AI & Data Engineer
Bilingual (English/Japanese)
- 15+ years experience (Australia & Japan)
- AWS / Azure / Databricks / ServiceNow
- Built multiple document automation pipelines
- Experience in ETL, data modeling, OCR pipelines, LLM integration
I specialize in turning messy unstructured documents into clean, reliable structured data.
Contact
📧 nueki@manoriworks.com
🌐 Portfolio
👉 Book a quick call (15mins)