I will do any kind of model finetuning ,deployment as API endpoint


About this gig
I build OCR and document-extraction systems that work on the documents everyone else's pipeline chokes on: bad scans, handwriting, dense tables, 200-page PDFs, non-Latin scripts.
Most recently I took a production OCR system from 65.5% down to 1.0% character error rate and made it 6× cheaper to serve on traditional Mongolian script, a vertical writing system with almost no training data anywhere in the world.
WHAT I DO
Model finetuning from unusable labeling into production level predictions. Data cleaning , Hypothesis testing to get best result out of most models so it can be used in product with good prediction rate.
Document extraction scanned PDFs, forms, invoices, receipts, archives, handwriting clean JSON, CSV or database rows, delivered with a per-field accuracy report so you know exactly what it gets right and where a human still needs to look.
Models into production containerised GPU endpoints, CI/CD, pinned dependencies, a latency budget agreed before I start. Not a notebook you can't deploy.
Get to know Davaadorj M.
AI Engineer
- FromMongolia
- Member sinceNov 2021
- Avg. response time1 hour
Languages
English, Mongolian, Japanese
