Ridhanya M

Lead Developer of IDA Chatbot

• Architected and deployed a Fast API based generative AI chatbot (IDA), integrated with OpenAI, to handle financial document queries and data analysis.
• Implemented an end-to-end data pipeline with context caching to minimize response latency, and optimize user experience.
• Leading a team of three engineers, driving feature development, conducting code reviews, and providing technical mentorship.
• IDA was publicly showcased and recognized at SFVegas 2025, with product contributions highlighted across platforms.
• Showcased by Intain at SFVegas 2025
• Featured for AI innovation in structured finance
• Spotlighted in Intain’s product innovation series
• In-depth Medium article: IDA – The Virtual Analyst

POCs & R&D

• Automated Loan Tape Mapping: Developed AI-based mapping of user-uploaded loan tape fields to ESMA-compliant standard fields.
• Collaboration & Innovation
• Collaborated with marketing and project management teams to align AI solutions with business objectives.
• Provided technical documentation, user training, and ongoing support for deployed AI tools.

• Co-developed Payables V3, an admin-processor web application for invoice processing that significantly improved business efficiency
• Led a proof-of-concept for EXI Service using CRF-based invoice digitization, achieving 80% accuracy
• Built dynamic NER extraction models to identify legal terms and created a chatbot-like interface for document uploads and field extraction
• Trained AI models to extract information from COVID-19 test reports with 85% accuracy

• Contributed to the IN-D Insurance project by developing NER models that mapped medical terms to ICD and CPT codes, aiding in automating insurance claims
• Part of the Payables solution team, leveraging AI/ML and rule-based methods to extract invoice data including table data with 85–90% accuracy

• Implemented IFRS 9 compliance report generation using Spring Framework for data validation, processing, and output generation
• Developed 30+ unit/integration tests following industry-standard methodologies to ensure code reliability

• Built a Python-based solution using OpenCV to detect private hire vehicles in apartment complexes, achieving 90% accuracy
• Designed region-of-interest extraction and license plate recognition algorithms to streamline parking management