Blue Machines AI launches Aurora speech-to-text model for BFSI

Blue Machines AI has launched Aurora, a multilingual speech-to-text model specifically engineered for the Banking, Financial Services, and Insurance (BFSI) sector in India. The model is designed to handle real-time conversations that involve Indian English, Hindi, Hinglish, and other code-mixed speech, even over noisy or low-bandwidth telephone connections.

Internal benchmarking indicates that Aurora achieved a Semantic Word Error Rate (WER) of 1.51% for English and 2.43% for Hindi BFSI conversations, with a 5.52% rate across multilingual speech. The model also recorded a BFSI Entity Error Rate of 4.23% for critical data points such as monetary amounts, interest rates, policy numbers, and transaction IDs.

Unlike general-purpose models, Aurora is trained to recognize specific financial terminology, including EMIs, KYC, premiums, SIPs, and NAVs. The company stated that the model is optimized for low-latency inference and high-concurrency environments to support banking, lending, insurance, and customer servicing workflows.