About the job
# Join Our Team
- The AI Platform team is on a mission to create a platform that enables everyone to use AI technology quickly and reliably. We technically support the use of AI across Toss.
- We are developing the necessary tools and platforms for new AI systems such as Retrieval-Augmented Generation, Agents, and Assistants to be rapidly experimented with and reliably operated.
- The platform we create is not just a simple toolkit; it is designed with scalability in mind, allowing AI technology to be implemented by more teams effectively.
- Addressing unresolved problems means that collaboratively defining and structuring technical directions is crucial.
- **Want to learn more about Toss’s data organization?** [→ *Toss Data Division Wiki*](https://recruit-data-division.oopy.io/)
# Responsibilities
- Integrate LLM-based components such as Retrieval, Generation, and Vector Search into a platform that various teams can reuse.
- Provide features that integrate both SaaS and self-hosted LLMs and ensure stable operations.
- Design the foundation for creating and experimenting with Agent systems more easily, including prompts, tools, and context configurations.
- After experiments, ensure stable operation in the service by systematizing and tooling the serving and operational flow of RAG and Agents.
- Develop a foundation to quantitatively evaluate the performance and quality of Agents and provide it as a platform.
- Design a common environment and experience to enable rapid experimentation and application of AI systems not only within the team but also across other teams.
- Structure unformatted technical elements and create directions that can expand into broader problems.
# We Are Looking For
- Individuals with experience applying technologies such as LLM, RAG, and Agents to real-world problems.
- Those who can technically define unstructured problems and systematically solve them.
- Candidates who have collaborated with multiple teams to develop and operate technology as if it were a product.
- Those who can quickly adapt to new technological trends and integrate them naturally within the team.
- Individuals interested in simplifying complex AI systems into a consistent and straightforward user experience.
# Preferred Qualifications
- Experience designing RAG components such as Retrieval, Generation, and Vector Search independently and integrating them at the system level.
- Experience selecting and operating various LLM serving structures (OpenAI API, HuggingFace, vLLM) tailored to service situations.
- Experience structuring various purpose-driven Agents and applying them in service operations is highly welcomed.
- Experience proactively designing experimental environments or tools based on the requirements of platform users (internal developers, model engineers, etc.).
- Experience creating and operating common platforms or RAG-based systems that can be extended across multiple projects or domains is especially desirable.
# Resume Tips
- If your past projects had significant impacts on the organization, please detail them.
- Rather than just listing languages, platforms, frameworks, or technologies used, provide context about the project's objectives, the methods you employed, and how you solved the problems.
- Include experiences where you resolved critical issues during platform operations or optimized performance/resources.
- If you have contributed to open source by fixing bugs or enhancing functionalities, please share those experiences.
# The Journey to Join Toss
- Application submission > Job interview > Cultural fit interview > Reference check > Compensation negotiation > Final acceptance
- The job interview will feature in-depth technical interviews focused on ML system design.
# A Message for Future Colleagues
“We don’t just serve rapidly evolving AI models; we build systems that ensure these models operate stably and are continuously improved.”
- The AI Platform team is responsible for the serving, experimenting, and operating infrastructure of various AI technologies such as LLM-based services, RAG systems, and search infrastructure to ensure smooth operation in production environments.
- We efficiently manage GPU resources and clusters, utilizing vLLM, Triton, Model Registry, etc., to automate experiments and deployments.
