Challenges

Limitations of Data Utilization in Global D2C Business

We operate Online Brand Shops (OBS) in over 40 overseas subsidiaries, developing our D2C business.
While vast amounts of data are collected on the OBS platform,
many overseas subsidiaries lack the specialized personnel and platforms to perform data-driven tasks.
Some countries face restrictions on data management and analysis due to compliance regulations.
Therefore, a platform that can analyze data and develop AI models while complying with local regulations was needed.

Solutions

AI-Based Customer Insight Automation for Overseas Subsidiaries

Digital transformation for overseas subsidiaries can now easily begin with Mellerikat.
The MLOps & LLMOps platform from Mellerikat automates the D2C digital transformation of overseas subsidiaries and
enables efficient production and operation of AI services while complying with local regulations. From generating inference-based insights to operation,
it creates an environment where even non-experts can easily provide AI services.

Core of AI Services: TCR-Based AI Modeling (MLOps)

Easy Index Model Creation with TCR AI Contents

Mellerikat provides various forms of AI contents based on the expertise of diverse data scientists.
Among them, TCR allows for the generation of various indices representing customer characteristics based on labeled customer states or behaviors in tabular data.
For example, by selecting various information such as customer purchase history and behavior patterns as inputs and specifying the desired label column, an index model can be easily created.

Development of Customer Characteristic Inference Indices Using D2C Customer Data

Analyzing customer characteristics by inflow channel and customer journey based on customer data collected from OBS is very important.
Mellerikat analyzes customer characteristics by inflow channel such as search, social media, and advertising,
as well as by customer journey stages such as product detail pages, adding to cart, and purchasing,
allowing for the easy development and operation of hundreds of AI models that can be utilized in digital marketing,
such as purchase likelihood indices and subscription conversion indices.
Additionally, the developed AI models are stably operated and managed through Mellerikat's MLOps platform,
efficiently supporting the entire cycle of model deployment, monitoring, and performance improvement.

Expanding with LLM: LLMOps for Overseas Subsidiary Users

Synergy of Domain-Specific AI and Generative AI

Generative AI (Gen AI) has versatility that can be utilized in various situations,
but in actual business, domain-specific AI tailored to each industry and purpose provides greater value.
Mellerikat supports the effective resolution of complex problems faced by companies by combining domain-specific AI with Gen AI.
This allows for a deep understanding of expertise and data patterns in each field, achieving higher quality predictions and analyses to maximize business value.
We provide four chatbot services based on Large Language Models (LLM) to maximize the business value of OBS's business areas.
These services are designed to make it easy for employees at overseas subsidiaries, even those with limited experience in digital transformation (DX), to utilize inferential data through natural language.

Intuitive Interpretation Support for AI Inference Results

The TCR index model provides customer-specific index values and the influence of the columns that
had the greatest impact on predictions through XAI (Explainable AI).
However, interpreting this information can be difficult for those who are not data experts, so the LLMOps platform converts inference results and
XAI information into natural language that domain experts can easily understand.
This enables effective utilization of AI results in practical fields such as marketing.

Marketing Support Based on Conversational AI

By combining indices generated in MLOps with logic based on LLMOps, we provide conversational AI services.
Domain experts can easily query data and receive practical solutions based on the inference results of index models.
In particular, we offer conversational AI functionalities for marketing DX that allow the easy extraction of target customer lists
based on inferential data, which can then be seamlessly communicated to marketing personnel.

Operations Based on MLOps and LLMOps Platforms

Automation and Efficiency of AI Development, Deployment, and Operation

By utilizing TCR AI Contents, we have reduced the AI model development period from 4 weeks to 2 weeks.
LLMOps allows for the API of various LLM logics to be created with simple commands, enabling rapid deployment and testing.
Developers can focus on LLM logic development without configuring backend infrastructure, greatly enhancing development efficiency.
LLMOps automates the entire process from Docker image building to cloud deployment, allowing for rapid service launches.

Data Security and Compliance Adherence Based on Edge App

AI models can be operated while complying with regulations even in countries with strict data regulations.
After anonymizing personal identifiable information, models are developed and deployed in the form of Edge Apps on local GCP.
Data does not leak externally and inference is done locally, providing AI-based insights while adhering to regulations.