Product descriptions:
Miridih provides an integrated design ecosystem that makes it easy for anyone to design and create. Their flagship service, MiriCanvas, enables nonprofessional users to create designs through an intuitive, template-based UX. Finished products can be seamlessly ordered via BizHows, which supports the production of various print materials such as business cards, stickers and more. With this robust design ecosystem, MiriCanvas has grown rapidly, surpassing 16 million cumulative subscribers and 13 million monthly design downloads in 2024. However, as data volumes increased, Miridih’s legacy architecture became inefficient, with structural limitations that placed a heavy burden on a small data analytics team. By adopting the Databricks Platform to modernize their data architecture, Miridih enhanced data literacy across the organization with improved data accessibility. The company has successfully embedded an autonomous, flexible, data-driven culture where most employees can retrieve and utilize data independently.
Legacy architecture was inefficient for large-scale data operations
With an integrated platform that supports the entire process from design to production, Miridih empowers creators by simplifying how ideas are brought to life. The company delivers a richer design experience by seamlessly connecting various services — including MiriCanvas and BizHows — into a unified ecosystem. As usage of MiriCanvas grew rapidly, so did the volume of online behavioral and service data. This created the need for a modern data analytics environment that could collect and analyze large-scale data in near real time and turn customer feedback into actionable product improvements.
However, Miridih’s legacy architecture relied on disjointed data pipelines and an inefficient computing environment, making data operations slow and complex. The fragmented, purpose-specific tools obscured data flow visibility, made maintenance difficult and created a steep learning curve for new analysts and nontechnical users.
They also lacked a robust governance framework for ensuring data quality and secure access. The rigid table structure made schema changes difficult and inflexible, often leading to missing data, inconsistencies and query errors. This, in turn, decreased trust in analytic outputs. Basic quality controls, such as managing permissions and tracking data flows, were hard to implement, delaying root cause analysis when failures occurred. On the infrastructure side, limited scalability resulted in frequent delays and repeated query failures.
As a result, key departments such as marketing and customer service struggled to make timely, data-driven decisions, and organization-wide self-service analytics was out of reach. “Our existing architecture struggled to keep up with the data demands of our rapidly growing business environment and had structural limitations across the collection, storage, analysis and governance of large volumes of data,” Miridih’s data engineer Jun Soo Lee explained.
Recognizing the need for a complete architectural overhaul, Miridih adopted the Databricks Data Intelligence Platform, which enables the management of large-volume data through a unified, scalable solution.
Redesigning the architecture completely with the Databricks Platform
Miridih fully re-architected their data platform with Databricks at the core. The company migrated to a Databricks serverless environment on AWS, integrating with AWS Glue and optimizing Spark code to improve data conversion and loading performance. For example, Lakeflow Jobs and clusters were implemented to streamline the management of customer online data, resulting in faster, more efficient pipelines for data ingestion and processing.
To centralize and secure metadata management, they also leveraged Unity Catalog, applying column masking and permission controls to protect sensitive information and enforce role-based access. Audit logs and table histories provided visibility into user activity, meeting key requirements for data protection and compliance.
The implementation of a medallion architecture, layered on Unity Catalog, introduced a structured approach to managing data quality. This model not only reduces unnecessary resource usage but also ensures consistent delivery of reliable analytics-ready data. Additionally, Miridih optimized performance and cost efficiency by separating cluster operations based on task type, improving overall data processing stability and scalability.
Databricks Notebooks allows teams to analyze data collaboratively through integrated code, visualizations and narrative context, making it easier to align across functions. Previously, collaboration was hindered by siloed analytics code stored on individual PCs or internal servers. Now, teams work together in a centralized, cloud-based environment that supports real-time collaboration and feedback.
The introduction of AI/BI Genie further empowered Miridih’s organization. Nontechnical users can now query data and gain insights independently, enabling fast and autonomous data-driven decision-making across departments. “Databricks is not just a technology adoption, but a key pillar in changing the culture of data utilization across the organization,” Jun Soo said.
Dramatically improved data literacy and utilization across the corporation
By leveraging Databricks to optimize their data pipeline, Miridih reduced the time required to collect and convert customer behavior data from over four hours to under one hour — a performance improvement of more than 75%. As a result, teams can now make data-informed decisions earlier in the day, shifting key strategy execution from the afternoon to the morning.
Since adopting Databricks, data literacy has significantly increased across the organization, creating a data-driven, autonomous work culture. Where data requests were once centralized with analysts and engineers, employees across departments — including designers, project managers and marketers — now use Databricks Notebooks and Genie AI to perform tasks such as querying, coding and visualizing data. Even non-data experts can now create their dashboards and use workflows for batch tasks.
Within just five months, more than 70% of Miridih’s workforce (approximately 280 employees) had adopted Databricks in their daily work, with over 200 dashboards and 180 workflows currently in use. Additionally, real-time visibility into resource usage has increased cost-awareness across teams, supporting more efficient operations.
“An environment where all members can use the data themselves has not only reduced the workload of the data engineers, but also increased the speed and accuracy of collaboration. We’ve successfully embedded a flexible, autonomous, data-driven culture throughout the organization," Jaeseung Ko, Data Platform Team Lead, explained.
This enterprise-wide transformation was not merely the result of adopting a new tool — it was the outcome of re-architecting the data foundation with Databricks and making data access intuitive and inclusive for all employees. The ability to automate tasks without constant technical team intervention has reduced the burden on data engineers and empowered teams to explore and act on data insights quickly.
Going forward, Miridih plans to expand their use of Databricks as a strategic foundation for AI development. The company aims to build a systematic MLOps environment by using MLflow to manage offline training and monitor the inference performance of deep learning models utilized in their search tools. Miridih is also actively developing an LLMOps system — focused on AI agent tracking, prompt governance, AI playground environments and LLM evaluation — leveraging MLflow to support these capabilities.
Miridih looks forward to continuing their collaboration with Databricks to accelerate data-driven innovation and scale their impact in the AI space.