AI research and OCR

I started in computer vision, reading curved shop signs, tiny microchip markings and photos taken almost in the dark. These days I work on financial paperwork.

Latest paper
PRISM · EMNLP 2026
Doctoral research
Low-light OCR
ICText dataset
10,000 images
PRISM result
FinLLM Poster Paper Award

Financial retrieval

2025 to 2026, AI Lens Group. EMNLP 2026 Industry Track

PRISM finds answer pages in long financial filings, such as last year’s operating margin. No training: a tuned system prompt, question-specific worked examples, retrieve–rerank–refine and one LLM listwise ranker. We measure accuracy, latency and cost. Our third-place ACM ICAIF 2025 entry grew into PRISM, winner of the FinLLM Poster Paper Award at IJCAI 2026.

PRISM pipeline: offline prompt library and exemplar store, prompt engineering, retrieve-rerank-refine, ranking preparation, a single LLM listwise ranker, and ranked output

Bank statement AI

2025 to 2026, AI Lens Group. ICLR 2026 workshop; earlier version an oral at ACM ICAIF 2025

AI-BAAM scores Malaysian MSMEs with limited credit history from six months of bank statements. It extracts account details and all transactions, checks PDFs for tampering, maps money flows and turns cash flow into credit-model features. Loan officers remain in the loop.

AI-BAAM pipeline: bank statement ingestion, key information extraction, transaction table extraction, fraud analysis, network analysis, and cash flow analysis feeding a credit scoring model, with loan officers in the loop

Low-light OCR

Doctoral research, University of Malaya. Publications: 2022–2025

Brightening alone can leave text unreadable. For my Ph.D., we built Text in the Dark, the first extremely low-light text dataset: 3,075 images and 59,975 annotated text instances. Our method restores text and brightness.

Low-light scene examples with text outlined, from the Text in the Dark dataset

Microchip text recognition

2020 to 2024, ViTrox and University of Malaya. ICDAR 2021 competition; Pattern Recognition

The markings on a chip are tiny, low-contrast and often scratched or broken, and factories need to read and inspect every one. With ViTrox, Baidu and partner universities we built ICText, the largest dataset for it, with 10,000 images and 100,152 characters annotated, and ran it as a competition at ICDAR 2021.

Close-up of a microchip with each printed character boxed and labelled

Conference publications

From text in difficult images to finding evidence in financial filings.

All conference papers

  • EMNLP 2026, Industry Track. PRISM: Prompt-Refined In-Context System Modeling for Financial Retrieval. Read paper Open code
    • Poster Paper Award, FinLLM Symposium @ IJCAI 2026.
    • Oral, 9th Workshop on Machine Learning in Finance @ KDD 2026.
  • ICIP 2023. Rethinking Long-Tailed Visual Recognition with Dynamic Probability Smoothing and Frequency Weighted Focusing. Read paper Open code
  • ICPR 2022. Extremely Low-Light Image Enhancement with Scene Text Restoration. Read paper Open code
  • ICDAR 2021. Competition on Integrated Circuit Text Spotting and Aesthetic Assessment. Read paper Open dataset
  • CVPR 2020. On the General Value of Evidence, and Bilingual Scene-Text Visual Question Answering. Read paper Open dataset
  • ICDAR 2019. Robust Reading Challenge on Arbitrary-Shaped Text (RRC-ArT). Read paper Open dataset
  • ICDAR 2019. Competition on Large-Scale Street View Text with Partial Labeling (RRC-LSVT). Read paper Open dataset
  • ACM MM 2018. Unprecedented Usage of Pre-trained CNNs on Beauty Product. Read paper

Papers and collaborators

All journal papers

  • Signal Processing: Image Communication, 2025. Text In the Dark: Extremely Low-Light Text Image Enhancement. Read paper Open code
  • Pattern Recognition, 2024. When IC Meets Text: Towards a Rich Annotated Integrated Circuit Text Dataset. Read paper Open code

Workshop papers

  • ICLR 2026 workshop. AI-BAAM: AI-Driven Bank Statement Analytics as Alternative Data for Malaysian MSME Credit Scoring. Read paper
    • The 2nd Workshop on Advances in Financial AI @ ICLR 2026.

Research collaborators

  • Earlier AI-BAAM: Cash Flow Underwriting with Bank Transaction Data: Advancing MSME Financial Inclusion in Malaysia (Read paper). Oral, AI for Financial Inclusion, Risk Modeling & Resilience in Emerging Markets @ ACM ICAIF 2025.
Chun Chet Ng, researcher in applied AI and computer vision