T

Toss

Fintech super-app for payments and banking

ML Engineer – OCR

Computer VisionConfirmed live in the last 24 hours
Full-timeSeoul, South KoreaNo salary listedIn-person

About the Job

Requirements

  • Experience applying OCR or Document AI models in production is desired.
  • Deep experience with at least one of two-stage OCR (detection plus recognition), end-to-end OCR, or VLM-based document understanding, with the ability to explain each method's advantages and disadvantages, is desired.
  • Experience with remote identity verification (eKYC), identity-document verification, or document forgery detection is an advantage.
  • Ability to design or modify vision model architectures and loss functions for a problem and optimize performance is desired.
  • Experience owning the full model lifecycle, from data collection, labeling, and synthetic data generation through training, serving, and operational monitoring, is desired.
  • Ability to understand financial-domain data involving identity documents and personal information, and account for masking, deletion, and access-control policies as technical design constraints, is desired.
  • Ability to translate ambiguous business requests into model problems and explain outcomes through business impact, such as cost savings or reduced lead time, is desired.
  • Experience fine-tuning or evaluating vision-language models for document understanding or information extraction is preferred.
  • Experience replacing external solutions with in-house models while achieving equivalent or better performance is preferred.
  • Experience improving generalization across data with different distributions, such as multiple countries or document types, is preferred.
  • Experience optimizing GPU serving with Triton, TensorRT, or ONNX, or improving throughput and cost through quantization or model compression, is preferred.
  • Experience designing and operating model-evaluation benchmarks or automated evaluation harnesses is preferred.

Responsibilities

  • Develop and improve text detection and recognition, document layout analysis, and vision-language-model-based end-to-end OCR models.
  • Develop models for key information extraction from financial documents such as identity documents, contracts, and government documents; document classification; document authenticity verification, including tampering and copy detection; and stamp matching.
  • Design generalized models for global eKYC that reduce dependence on country-specific training data, including key information extraction, quality checking, and fraud detection, and build reusable training, evaluation, and deployment pipelines.
  • Define OCR and Document AI problems and establish data-collection strategies, including synthetic data generation, labeling strategies, and personal-information masking policies.
  • Build the company's Document AI benchmark and continuously evaluate new models and vision-language-model releases for accuracy, speed, hallucination, and cost to inform model adoption.
  • Validate consistency between training and serving, deploy models in a GPU serving environment using Triton, and diagnose and improve production-model performance issues using real data.

About the Company

Toss

Fintech super-app for payments and banking

Toss is a South Korean financial services platform that started as a peer-to-peer money transfer service and has grown into a digital 'super app' for personal finance. Its products are accessed through one mobile interface that unifies bank accounts, credit cards, loans, investments, and other financial services, including its own Toss Bank and Toss Securities. It differentiates itself by offering a broad, integrated ecosystem—combining digital banking, asset management, and brokerage services under one roof and running its own banking and securities units. Its goal is to become the leading consumer finance platform in South Korea while expanding its user base globally.

Company size
1,001-5,000
Company stage
Series G
Total funding
$1.5B
Headquarters
Seoul, South Korea
Founded
2011

Recently Posted Jobs

Explore more opportunities at Toss

See more jobs