SARADUNE6 CHOUDHARY

Developed and deployed a multilingual AI chatbot using LangChain, RAG for contextual responses, and Devtron for scalable deployment. Designed and deployed a multilingual AI chatbot for the FinTech domain, enabling secure and context-aware interactions in multiple languages. Leveraged LangChain for conversational workflows and RAG for precise, domain-specific responses. Built robust data pipelines and customized embeddings for financial accuracy and multilingual support. Deployed with Devtron, ensuring a scalable and compliant production solution. Enhanced client accessibility and support, reducing response times by 40% and boosting customer satisfaction by 30%, contributing to increased user retention and engagement.

Developed scalable solutions for name matching and loan propensity prediction to enhance merchant onboarding and risk assessment. Developed and optimized a BERT-based name-matching model for verifying merchant identities, reducing manual checks and increasing KYC accuracy by 15%. Used BERT, Siamese BERT, and FuzzyWuzzy for improved name matching accuracy, handling common issues like spelling errors and transliterations. Deployed real-time processing for automated KYC workflows, integrating models into production systems and reducing verification time by 20%. Developed and deployed a XGBoost model predicting the tendency of a merchant to take up a loan on a given day. Automated daily scoring for 4 lakh merchants, using a on-us feature set based on merchant attributes and time. The model segregates merchants into propensity tiers, delivering a KS score of 0.49 with an AUC of 80-85%. The model helps achieve a 6% incremental gain the loan-lending funnel for the organisation through data-driven decision-making for targeted loan sales. Developed an interoperable speaker notification system for BharatPe, using NLP classification and NER models (e.g., Mistral, GLiNER) to replace error-prone regex. Enhanced accuracy by implementing smart text classification and entity extraction, reducing missed or incorrect alerts and improving system reliability.

Streamlined financial evaluations and loan approvals through advanced data analytics and risk assessment. Analyzed Daily Payment Data (DPD) to assess merchant financial health and risk, identifying trends to support loan risk assessments. Queried large datasets to ensure data accuracy, compliance, and performance insights for strategic planning. Built and optimized data pipelines for efficient processing, enabling real-time insights for business strategies. Evaluated merchants' financial health and risk levels using DPD, automating risk segmentation for high-risk and low-risk classification. Designed dashboards to visualize trends and support strategic decisions. Presented insights to leadership and collaborated with NBFCs to identify merchants for loan approvals. Built models to assess merchant creditworthiness, improving compliance and precision.