Learning from a ₹6 Share to Build FinEd Saathi
A developer shares a mysterious trading loss on a ₹6 share that inspired the creation of FinEd Saathi, a voice-first financial literacy tutor built in 10 days.

Stock photo for illustration only, not from the actual event
- Developer built FinEd Saathi during the 10 Days of Voice Agents challenge
- Addresses the lack of basic knowledge regarding brokerage and investment taxes
- Powered by Deepgram, Gemini, and Murf Falcon 2 speech models
- Includes a paper trading portfolio starting with ₹1,00,000 in virtual cash
The origin of this project began when the developer purchased a share priced at roughly ₹6 without fully understanding the total transaction costs. After selling it around the same price, a mysterious loss of about ₹50 appeared in the account without knowing whether it came from the contract note, ledger, or P&L view. The crucial lesson was not that every small trade incurs the exact same fee, but rather that trading occurred without any fundamental grasp of brokerage, taxes, or the origin of figures displayed inside a broker application.
That exact experience served as the foundation for FinEd Saathi, a voice-first financial literacy tutor built during the 10 Days of Voice Agents challenge under the VoiceForBharat Edition. The developer chose the Financial Services track to solve a personal pain point, deliberately merging finance with education because the root problem was financial access without comprehension. The AI tutor targets beginners seeking patient explanations of Indian market concepts before risking real capital.

Stock photo for illustration only, not from the actual event
Financial education often presumes learners already understand complex terminology, sending novices straight to dense fee schedules or tax circulars. FinEd Saathi was designed to meet learners where they are, explaining stocks, mutual funds, SIPs, ETFs, gold, F&O, IPOs, and bonds while untangling confusing charges without making blind guesses. In the case of the ₹6 share story, the agent treats profit or loss as zero until the user points out where the discrepancy appeared.
Voice-driven financial applications require exceptionally low latency to maintain a natural conversational flow. When users ask complex questions about monetary concepts, long pauses can instantly make the interaction feel robotic or unnatural, highlighting why ultra-fast text-to-speech models are critical for educational voice agents.
The technical architecture leverages Deepgram Nova-3 for multilingual speech recognition, Gemini for conversational reasoning, and LiveKit for real-time streaming sessions. Murf Falcon 2 provides the voice using Nikhil, delivering an Indian conversational tone with a remarkably low time to first audio of roughly 100 milliseconds. Additionally, a paper portfolio feature lets beginners practice safely with ₹1,00,000 in virtual cash without linking a real brokerage account.
Source: Dev.to
Found something wrong in this article? Report an issue with this article
Comments
Leave a Comment