-Legal Entity: Hyperconnect -Brand: Match Group AI
About Match Group AI Match Group AI is a centralized technology organization that solves core online dating challenges using AI. We focus on key areas of the user experience, such as profiles, matching, and Trust & Safety, to identify and tackle complex problems. By applying the latest AI technologies and data-driven approaches, we innovate the user experience. Furthermore, we expand the group's common technical foundation through collaboration with global dating brands within Match Group, such as Tinder and Hinge.
If you would like to learn more about the projects our team is working on, please check out the following post: ๐ "Introducing the Match Group AI Team."
About the Match Group AI ML Team The ML Team consists of ML Engineers who apply AI/ML technologies to various Match Group services. Our work begins by identifying and defining problems that arise during the creation and operation of actual products. We develop or reproduce the most suitable State-of-the-Art (SotA) models for problem-solving and deploy the finished models reliably and efficiently into mobile and server environments. Through continuous monitoring and improvement, we build the AI Flywheel for our services. In this process, we collaborate closely with various specialized organizations, including backend/frontend/DevOps engineers, data analysts, and PMs, to create AI experiences that make a real impact on users. For more details on how we work, please refer to the following:
- [How AI Lab Works] Head of AI - Shurain Interview
- AI in Social Discovery(Blending Research and Production)
Some of our work is shared externally through research papers or open-source code. When building ML models for product use, existing research is often insufficient. To fill these gaps, project participants collaborate to refine the meaningful parts of the research and, whenever possible, release the results along with the code. As a result, we have achieved approximately 20 external research accomplishments, including the following:
- 2024 CUPID: Real-time Session-based Mutual Recommendation System for 1:1 Social Discovery Platforms, presented at ICDM Workshop
- 2023 TiDAL: Active Learning based on Model Behavior during Efficient Training, published at ICCV 2023
- 2023 Research on setting thresholds to satisfy multiple classification criteria in moderation environments, published at WSDM 2023
- 2022 Research on increasing semantic diversity in dialogue generation, published at EMNLP 2022
- 2022 Methods for effective learning in environments with severe label noise, published at ECCV 2022
- 2022 Research on chatbots that mimic a target character using only a few utterances, published at NAACL 2022
- 2022 Research on improving performance in dialogue generation models using examples, presented at ACL 2022 Workshop
- 2022 Research on distillation techniques for audio classification in mobile environments, published at ICASSP
For ML research to progress effectively, the infrastructure for deep learning training must be well-equipped. We maintain our own deep learning cluster (total of 160 A100 GPUs and 40 H100 GPUs) to ensure ML engineers can fully develop and experiment with models. Additionally, we build and operate our own data pipelines, including data collection and preprocessing, using cloud services. We also work alongside various software engineers (backend/frontend/DevOps/MLSE) who help bring ML models to production.
Responsibilities Match Group AI's ML Engineers are both scientists who research and apply the latest AI/ML technologies and engineers who design and optimize models and systems for real-world service environments. Our organization handles a variety of tasks, and we are looking for individuals who are proficient in at least one of the following areas:
- Solving various business problems that arise in products serviced by Match Group. This involves understanding the background, goals, and constraints of business problems, redefining them as correct AI/ML problems, and finding the best methodology to solve them.
- Participating in the development of new products and features for Match Group. This involves utilizing AI/ML technologies to quickly and correctly implement products and features, from ideation and prototyping to reaching actual users.
- Conducting advanced research and development on large language models and multimodal models. This involves using Match Group's data and the latest technologies to create models specialized for the dating domain and finding ways to apply them to various areas of the product.
To perform these tasks, we expect Sr. ML Engineers to have a solid foundation in AI/ML and the ability to lead the team's technical decision-making based on rapid acquisition of the latest technologies. Based on these capabilities, you will primarily lead the following technical topics:
- Methods for correctly handling data with various modalities, including text, images, and event logs, and methods for correctly resolving bias and noise in collected data.
- Defining evaluation metrics aligned with product goals and optimal model training methods to achieve those goals.
- Methods for optimizing on-device and large-scale model inference, considering engineering constraints and infrastructure environments.
Requirements
- Individuals with basic knowledge of AI/ML, in-depth knowledge of at least one specific domain, and relevant project experience.
- Individuals who can discover statistical characteristics and patterns in data through Exploratory Data Analysis (EDA) and apply them to the problem-solving process.
- Individuals proficient in Python and deep learning frameworks such as PyTorch, TensorFlow, or JAX, capable of writing high-quality code that is collaborative and maintainable.
- Individuals with engineering skills required for building and deploying ML model training pipelines.
- Individuals with a strong interest in the service-oriented application of AI technology.
- Individuals who consistently learn new technologies and share them with the team.
- Individuals capable of smooth communication in Korean, and who can understand/write technical documents in English and communicate remotely.
Preferred Qualifications
- Individuals with publication records in top-tier AI/ML conferences and journals (NeurIPS, ICLR, ICML, ACL, RecSys, KDD, CVPR, etc.) or awards in relevant competitions.
- Individuals with experience in significantly improving key business metrics by applying AI technology to actual services.
- Individuals who can discover meaningful insights from data and apply them to decision-making using statistical techniques and SQL-based analysis, such as A/B test planning, target KPI definition, and causal analysis.
- Individuals with project development experience outside of the AI field, including client-side (Android, iOS) and backend.
- Individuals with experience in defining requirements, managing schedules, and adjusting priorities at the project or feature level while collaborating with various non-engineering departments (PM, data, business, etc.) (including PM/PO experience).
- Individuals capable of technical communication in English during large meetings or discussions involving multiple stakeholders.
Hiring Process
- Employment Type: Full-time
- Hiring Process: Document Screening > Pre-assignment > Live Coding Interview > Technical Interview > In-depth Technical & Cultural Alignment Interview > Final Offer (* The process may be subject to change if necessary.)
- Document screening results will be notified individually to successful candidates.
- Application Documents: Detailed career-based English resume (PDF) in free format.
- This position is eligible for the Professional Research Personnel (military service exemption) program. For those under the military service exemption, service management will be conducted in accordance with relevant laws.
About the Match Group AI ML Team The ML Team consists of ML Engineers who apply AI/ML technologies to various Match Group services. Our work begins by identifying and defining problems that arise during the creation and operation of actual products. We develop or reproduce the most suitable State-of-the-Art (SotA) models and deploy the finished models reliably and efficiently into mobile and server environments. Through continuous monitoring and improvement, we build the AI Flywheel for our services. In this process, we collaborate closely with various specialized organizations, including backend/frontend/DevOps engineers, data analysts, and PMs, to create AI experiences that make a real impact on users. For more details on how we work, please refer to the following:
[How AI Lab Works] Head of AI - Shurain Interview AI in Social Discovery(Blending Research and Production)
Some of our work is shared externally through research papers or open-source code. When building ML models for product use, existing research is often insufficient. To fill these gaps, project participants collaborate to refine the meaningful parts of the research and, whenever possible, release the results along with the code. As a result, we have achieved approximately 20 external research accomplishments, including the following:
2024 CUPID: Real-time Session-based Mutual Recommendation System for 1:1 Social Discovery Platforms, presented at ICDM Workshop 2023 TiDAL: Active Learning based on Model Behavior during Efficient Training, published at ICCV 2023 2023 Research on setting thresholds to satisfy multiple classification criteria in moderation environments, published at WSDM 2023 2022 Research on increasing semantic diversity in dialogue generation, published at EMNLP 2022 2022 Methods for effective learning in environments with severe label noise, published at ECCV 2022 2022 Research on chatbots that mimic a target character using only a few utterances, published at NAACL 2022 2022 Research on improving performance in dialogue generation models using examples, presented at ACL 2022 Workshop 2022 Research on distillation techniques for audio classification in mobile environments, published at ICASSP
For AI research to progress effectively, the infrastructure for deep learning training must be well-equipped. To ensure ML engineers can fully develop and experiment with models, we operate a deep learning cluster based on AWS, consisting of 10 DGX nodes (8 H100 GPUs per node, total of 80 H100 GPUs). Additionally, we build and operate our own data pipelines, including data collection and preprocessing, using cloud services, and collaborate with various software engineers (backend/frontend/DevOps/MLSE) who help bring ML models to production.
If any information submitted is found to be false or if there are grounds for disqualification under relevant laws, the offer may be rescinded. If necessary, additional screening and document verification may be conducted beyond the previously announced hiring process. National veterans will be given preference in accordance with relevant laws; please notify us when applying if you are eligible and submit supporting documents upon hiring. When applying for a position at Hyperconnect, this Privacy Policy applies regarding the processing of personal information: https://career.hyperconnect.com/privacy
#HPCNT