Case Study
Generative AI & ML

Developed an AI-powered Smart Learning Assistant to Enhance Information Retrieval for a Financial Planning Platform

About This Project

The customer observed a growing trend of user queries beyond platform-related inquiries. Users sought conceptual understanding, terminologies, compliance requirements, and more. To address this, our customer aimed to provide an interactive and automated learning assistant to enhance financial literacy among users.


Generative AI & ML


About the Client

Our customer is a digital financial planning app based in Malaysia. It offers end-to-end solutions and helps users execute their financial plans efficiently.

Understanding the Challenge

The customer reached out to us with the aim of developing an AI learning assistant capable of understanding complex financial queries and providing personalized recommendations. The customer started their own financial literacy program, which can be accessed via their website. 

However, since data kept on growing, covering every aspect was not possible, and so they wanted a smart learning assistant for their customers. The existing search systems struggled to efficiently respond to heterogeneous data (CSVs, PDFs, images, videos) and simplify complex finance concepts for users. 

The platform currently catered to 11k daily active users which was anticipated to grow significantly. Additionally, the evolving knowledge base demanded an AI solution that could adapt to incremental data ingestion without heavy retraining. The solution needed to tackle the following challenges seamlessly:

  • Handle Complexity: Users sought explanations on various financial concepts, terminologies, compliance regulations, and procedural guidance. These queries ranged from basic to advanced, requiring an AI system capable of understanding and responding appropriately.
  • Access Heterogeneous Data: Relevant information was scattered across various formats such as text, PDFs, CSVs, videos, audio, and webpages. Integrating and processing this data efficiently to provide coherent responses posed a significant technical challenge.
  • Stay Updated: Financial information is subject to frequent updates due to regulatory changes, market trends, and new developments. The customer needed a solution that could automatically update itself to provide the latest and most accurate information to users without frequent retraining.
  • Ensure Intelligibility: Simply providing access to information was insufficient; the solution needed to ensure that users could understand the content provided. This required not only accurate responses but also language that was accessible and comprehensible to users with varying levels of financial literacy.

Velotio's solution revolutionized our platform, enabling users to easily access relevant information and significantly improving their learning experience. Their expertise and dedication were invaluable in achieving our goals.

Head - Technology

How We Made It Happen

Velotio assigned a team comprising AI engineers, Python developers, and Frontend Developers to develop the solution in a span of 4 months. Before developing the smart learning assistant, the team initiated a comprehensive requirement-gathering process to gain a deep understanding of the customer's users and needs. 

Requirement Gathering -

  • This phase involved collecting valuable insights into user preferences, satisfaction levels, and interaction patterns with the existing systems. 
  • Gathered information on whether users preferred textual explanations, visual aids, interactive tutorials, or a combination of these methods.
  • Analyzed user feedback and engagement metrics to gauge how effectively the current system addressed user queries and provided relevant information.
  • Identified the knowledge gaps that users sought to fill, the effectiveness of the program in addressing these gaps, and any areas for improvement.
  • By gathering these requirements, the team gained valuable insights into the needs, preferences, and challenges faced by the users. This laid the foundation for developing an AI learning assistant that would effectively cater to these requirements, providing users with a seamless and enriching learning experience in finance.

MVP Development and Implementation -

  • Worked closely with the key stakeholders to design a minimal yet efficient data infrastructure capable of handling diverse data sources and ensuring seamless updates.
  • Developed a RAG-based approach and customized data indexing using the llama-index library.
  • Conducted research to build trust and gather early feedback.
  • Deployed the final solution using a tech stack comprising Python3, MongoDB, Hugging Face Transformers, Pytorch, Flask, AWS, EC2, CUDA, Nebula Graph DB, Llama-index, and Qdrant vector DB. Ensured seamless integration with the customer’s platform and provided the necessary support for ongoing maintenance and updates.
  • Implemented Custom Triplet Generation Logic - enhanced logic to retrieve dispersed data sources - enabling a comprehensive understanding of topics addressed indirectly across multiple formats.
  • Implemented automatic index expansion and updates for incremental data management, allowing efficient storage and retrieval of new data without heavy computing costs.
  • Employed Nebula Graph DB, Llama-index, and Qdrant vector DB for efficient indexing, storage, and retrieval of heterogeneous data.

How Velotio Made a Difference

Supports incremental data storage and indexing, saving 50%-70% of indexing time, that is, nearly 5000 EC2 (C5a.xlarge) compute hours for a dataset of ~450 English text documents.

Over 90% of users, quantified through a statistically significant sample size (~ >10K end-users), reported an intelligible learning experience, as measured by a Likert scale survey with a 95% confidence interval and a margin of error of ±5%.

The average CSAT increased from 3.5 to 4.2 out of 5.

With Velotio, achieve breakthroughs in your product development journey.

Over 90 global customers, including NASDAQ-listed enterprises, unicorn startups, and cutting-edge product companies have trusted us for our technology expertise to deliver delightful digital products.

Talk to us

Work with modern and scalable technologies

We leverage emerging technologies to build products that are designed for scalability and better usability.

Rated 4.6/5 on Clutch

325+ highly skilled engineers

With us as your tech partners, you get access to a pool of digital strategists, engineers, architects, project managers, UI/UX designers, Cloud & DevOps experts, product analysts and QA managers.

At Velotio, we hold ourselves to sky-high standards of excellence and expect the same from our customers.