Machine Learning Engineer (Match Group AI)
Match Group · Seoul, South Korea ·
- Work mode
- Onsite
- Category
- ML ai
- Visa sponsorship
- Yes
- Company size
- 501-1000
Build and deploy AI models for online dating apps like Tinder and Hinge, focusing on matching, Trust & Safety, and user experience using LLMs and multimodal data.
Match Group AI Team Introduction
Match Group AI (MG AI) is the central tech organization that drives innovation across Match Group's global portfolio, including Tinder, Hinge, Azar, Pairs, Match, BLK, and more. Match Group aims to spark meaningful connections for everyone, worldwide, and the MG AI team's role is to apply cutting-edge AI to some of the hardest challenges along that journey, across diverse domains (e.g., Recommendation, Trust & Safety, Profile Enhancement).
Unlike brand-specific teams (e.g., Tinder, HYPERCONNECT AI), the MG AI team offers a unique opportunity to impact the entire Match Group ecosystem. You won't just build for one app; you'll help develop scalable AI solutions that power Tinder, Hinge and beyond, defining the technological gold standard for the global dating industry.
- Detailed article: Introduction to Match Group AI Team (written in Korean)
Working as a Machine Learning Engineer at MG AI
ML Engineers at MG AI own the model and the metric it optimizes. You turn broad product goals and data into ML problems worth solving, build the models that address them, and measure their impact.
Your responsibility covers the decisions around the model as much as the model itself. You choose the optimization target that translates into business results, design how the training data behind that target is collected and processed, and set the online and offline evaluation criteria that determine whether the model solved the problem.
You will work across a wide range of ML problems, and the problems change as the products and the technology do.
Recent examples from the team:
- Utility modeling: Move beyond event prediction (whether a user will like another user, whether a user will report someone) to model the utility each event delivers to users and to the system, and optimize for the utility of the system as a whole.
- Cold start and data scarcity: Build models that work when data is missing or thin — prototyping stages, privacy constraints, rare events — and design the loop that collects the data those models need next.
- LLM and agentic systems: Leverage LLMs and agentic systems in Recommendation, Trust & Safety, and Profile Enhancement, both as products in their own right and as a way to raise the performance of existing models.
- Generative model evaluation: Define evaluation metrics for generative models that connect to business outcomes and that the model can be optimized against.
Our engineers also publish selected technical work on the Hyperconnect Tech blog (written in Korean).
- How to Set ML Objectives
- How Hyperconnect Built an LLM Explanation Policy
- LLM-as-a-Judge for Explanation Quality
$ - $ a year
None