Built core NLP and computer vision capabilities powering ValueFirst's conversational platform — from intent detection and entity recognition to document OCR pipelines and a unified ML API used across all product lines.
Surbo Natural Language Processing Engine
Devoted significant effort to developing natural language processing capabilities for the Surbo platform — the foundation for constructing enterprise chatbots. Built Named Entity Recognition (NER) and intent detection pipelines that powered customer-facing conversational flows across multiple client deployments.
The NLP stack handled multi-domain intent classification, slot filling, and context management, giving the Surbo platform the ability to understand domain-specific language without per-client fine-tuning.
Optical Character Recognition for Structured Documents
Developed a modular OCR system applicable across the company's product lines — architected to be interchangeable across different document types and use cases. The system was deployed for Domino's Indonesia's scan-and-win campaign, where high-throughput receipt OCR was required at scale with low error tolerance.
Text Enhancement & Unified ML API
Implemented multiple text improvement capabilities across products: a paraphrase API, a grammar correction module used for model retraining feedback loops, and data correction functionality with integrated spell-checking. All capabilities were surfaced through a unified ML API that gave every chatbot product organisation-wide access to the same underlying model infrastructure without duplication.