Raise Financial Services Adds Cartesia Voice AI for Multilingual Finance in India

Voice added to Raise's financial AI stack

Raise also runs the investing platform Millions and the AI-powered financial markets research assistant fuzz. According to Cartesia, the company has invested heavily in AI built specifically for Indian financial markets. Its team developed artham, a 7-billion-parameter small language model trained exclusively on Indian financial and capital markets data. It is built and hosted in India, and AWS recognized it at re:Invent 2025 in Las Vegas.

Artham has run inside fuzz, the stock screening platform ScanX and Dhan for over a year, summarizing market news, reading context into price moves and answering user queries with live data from official sources. Voice is the newest addition to that stack, and Cartesia supplies the speech models.

Alok Pandey, Raise's Co-Founder and Chief Technology Officer, said Cartesia provides "best-in-class voice models that integrate well with the real-time voice infrastructure that we've built in-house." Raise plans to scale the partnership across Dhan, Millions, the trader education platform Upsurge, the insurance distribution platform Pluto and other group entities.

Early results from Dhan's voice bot

Cartesia's models currently power a portion of Dhan's voice bot interactions, including conversational KYC. Human teams still lead on servicing, complex queries and relationship support.

• Over 75% of these interactions happen in languages other than English.
• Hindi alone accounts for more than half.
• Over a third of Raise's voice bot calls are handled outside business hours, entirely without agent involvement.

The out-of-hours figure points to users who reach out in the evenings and overnight, according to the source.

Technical demands of spoken finance

Cartesia says speaking financial information aloud brings its own requirements. A written answer must keep its meaning when spoken, numbers need an unambiguous pronunciation, and financial terms must fit naturally into the surrounding language. Audio generation also has to start quickly enough to sustain a conversation.

Sonic, Cartesia's text-to-speech model, offers a choice of voices and regional accents, with controls for pace and emotional expression. It supports 11 Indic languages, including Hindi, Tamil, Telugu and Marathi, and allows language code-switching within a single generation. This lets teams keep a consistent voice as a conversation moves between languages.

Cartesia also points to its work on transcript fidelity and locale-aware pronunciation, which covers alphanumeric identifiers, dates and monetary amounts. This matters when English financial terms appear inside an Indian-language sentence. According to Cartesia, its models are built for secure, high-volume, real-time workloads, with deployment options in India. Raise runs them within its own infrastructure, which is built to meet the data-handling and governance standards that financial applications require.

A design partnership

Raise has also been an early design partner and has contributed to Cartesia's multilingual model development. Cartesia argues that broad measures of naturalness do not on their own show whether a model preserves numerical detail in a code-switched financial sentence, and that domain expertise helps identify the cases that need closer evaluation.

For users, the early figures suggest that voice access in Hindi and other Indian languages, including outside business hours, is already in use at Dhan. Both companies say they share the aim of making financial information more accessible through conversation in the languages people use naturally. Cartesia says it will keep supporting Raise's voice development as the partnership expands.