We use cookies to run the site and count visits. "Got it" also allows session recording.
HYPERCONNECT
Seoul, South Korea2 months ago
Grade
Staff
Experience
from 5 years
Work format
Hybrid
Employment
Full-time
English level
B2, upper-intermediate
An email or a Telegram message to the recruiter
HYPERCONNECT AI is an organization that drives real business results through machine learning research. We reframe business problems as ML tasks and focus on improving key product metrics. We avoid simply applying off-the-shelf solutions, instead rethinking problem formulation and identifying the best approaches ourselves. As a result, HYPERCONNECT AI research has led to real business impact, with some of these cases featured in our parent company Match Group's earnings reports. At HYPERCONNECT AI, research is a means, not an end; our goal is clear business impact. Below are examples of problems we solve at HYPERCONNECT AI: Scaling Recommendation Models: While scaling laws are widely recognized in LLMs, this is not yet the case for recommendation models. We conduct research to create structures that improve performance as models and data scale. We work with large models, including LLMs, but design models with practical constraints such as inference in mind. Aligning Recommendation Models: All recommendation tasks ultimately boil down to quantifying 'how much will the business grow if this user is shown this content.' We explore methods for post-processing scalable recommendation models using reinforcement learning, causal inference, and game theory to align with business objectives. Automating ML Moderation Workflows with Multimodal LLMs: We conduct research to automate and enhance data labeling, model improvement, and evaluation steps in Trust & Safety (moderation) workflows that span multiple modalities, such as images, voice, and text, using LLMs. We focus on ensuring LLMs accurately understand internal company data and policies, and that their judgments are robust to the level of internal experts. Meaningful research outcomes are sometimes published as papers or open-sourced. We do not aim to publish papers, but if aspects arise during the course of solving product problems that cannot be fully explained by existing research, we systematize them and leave them as research outcomes (see references). To date, we have published around 20 external research outcomes. To solve business problems with AI, one must also have the proper infrastructure for training models. At HYPERCONNECT AI, we have built and utilize our own GPU cluster for research, allowing ML Research Scientists to fully focus on model development and experimentation.
Staff, Machine Learning Research Scientist (Azar) At Azar, a service that connects people, we tackle problems that are out of reach for existing technologies but solvable with machine learning. We solve problems in recommendations, moderation, and Azar's new innovative areas, aiming for the technologies we develop to contribute to the growth of the actual service.
As a key partner to the Head of AI, we identify critical problems to solve in Azar's recommendation and moderation domains. We align perspectives across various professions and leaders, and jointly make decisions on organizational priorities and implementation strategy based on business impact and technical feasibility. We identify opportunities that cannot be discovered from existing product requirements alone, drawing on the latest research and advancements in AI/ML. We propose how new technologies can revolutionize Azar and evaluate whether they are worth investing in, based on technical and business feasibility, as well as long-term technical advantages. We reframe business requirements as AI/ML problems and develop solutions, considering a wide spectrum of methods, including not only deep learning but also classical machine learning, statistics, and optimization. We do not limit ourselves to iteratively adding heuristics and exceptions to meet business requirements but, based on a deep understanding of problem structure and ML theory, seek directions close to theoretically optimal within given constraints. We directly tackle challenging ML problems and pressing business issues. We strive not only for offline model performance but also validate changes in user experience and real service metrics to ensure long-term product and business growth. We understand the entire ML system, connecting data, training, evaluation, experimentation, serving, and feedback, going beyond the performance of individual models. We design the structure of the entire system and the role of each component from an ML perspective to enable fast and reliable experimentation and continuous improvement. We align research themes and solution directions with other ML Research Scientists, guiding each research effort towards a unified technical strategy and product outcome. We collaborate with Machine Learning Software Engineers, Data Engineers, Product Teams, etc., to make technical decisions that lead to research outcomes being applied to real products.
PhD degree in Machine Learning or related fields, or equivalent research experience. 5+ years of experience in developing and improving ML-based products or services in the industry. Experience in defining clear ML problems from ambiguous business problems and developing solution strategies based on theoretical rationale and experimentation. Experience in independently and end-to-end solving difficult and challenging problems, and connecting model performance to improvements in user experience or business metrics. Ability to align technical directions and make decisions persuasively through collaboration with other researchers, engineers, and non-technical stakeholders.
Experience publishing papers as a primary author in top AI/ML conferences (NeurIPS, ICLR, ICML, CVPR, KDD, etc.). Research experience in key technologies such as recommendation systems, reinforcement learning, LLMs, and AI Agents. Experience in bringing ML systems to market and iteratively improving them for user-facing, high-traffic services. Ability to lead technical discussions with global teams in English.
Employment Type: Full-time
About HYPERCONNECT AI HYPERCONNECT AI is an organization that drives real business results through machine learning research. We reframe business problems as ML tasks and focus on improving key product metrics. We avoid simply applying off-the-shelf solutions, instead rethinking problem formulation and identifying the best approaches ourselves. As a result, HYPERCONNECT AI research has led to real business impact, with some of these cases featured in our parent company Match Group's earnings reports. At HYPERCONNECT AI, research is a means, not an end; our goal is clear business impact. Below are examples of problems we solve at HYPERCONNECT AI: Scaling Recommendation Models: While scaling laws are widely recognized in LLMs, this is not yet the case for recommendation models. We conduct research to create structures that improve performance as models and data scale. We work with large models, including LLMs, but design models with practical constraints such as inference in mind. Aligning Recommendation Models: All recommendation tasks ultimately boil down to quantifying 'how much will the business grow if this user is shown this content.' We explore methods for post-processing scalable recommendation models using reinforcement learning, causal inference, and game theory to align with business objectives. Automating ML Moderation Workflows with Multimodal LLMs: We conduct research to automate and enhance data labeling, model improvement, and evaluation steps in Trust & Safety (moderation) workflows that span multiple modalities, such as images, voice, and text, using LLMs. We focus on ensuring LLMs accurately understand internal company data and policies, and that their judgments are robust to the level of internal experts. Meaningful research outcomes are sometimes published as papers or open-sourced. We do not aim to publish papers, but if aspects arise during the course of solving product problems that cannot be fully explained by existing research, we systematize them and leave them as research outcomes (see references). To date, we have published around 20 external research outcomes. To solve business problems with AI, one must also have the proper infrastructure for training models. At HYPERCONNECT AI, we have built and utilize our own GPU cluster for research, allowing ML Research Scientists to fully focus on model development and experimentation.
Staff, Machine Learning Research Scientist (Azar) At Azar, a service that connects people, we tackle problems that are out of reach for existing technologies but solvable with machine learning. We solve problems in recommendations, moderation, and Azar's new innovative areas, aiming for the technologies we develop to contribute to the growth of the actual service.
Responsibilities As a key partner to the Head of AI, we identify critical problems to solve in Azar's recommendation and moderation domains. We align perspectives across various professions and leaders, and jointly make decisions on organizational priorities and implementation strategy based on business impact and technical feasibility. We identify opportunities that cannot be discovered from existing product requirements alone, drawing on the latest research and advancements in AI/ML. We propose how new technologies can revolutionize Azar and evaluate whether they are worth investing in, based on technical and business feasibility, as well as long-term technical advantages. We reframe business requirements as AI/ML problems and develop solutions, considering a wide spectrum of methods, including not only deep learning but also classical machine learning, statistics, and optimization. We do not limit ourselves to iteratively adding heuristics and exceptions to meet business requirements but, based on a deep understanding of problem structure and ML theory, seek directions close to theoretically optimal within given constraints. We directly tackle challenging ML problems and pressing business issues. We strive not only for offline model performance but also validate changes in user experience and real service metrics to ensure long-term product and business growth. We understand the entire ML system, connecting data, training, evaluation, experimentation, serving, and feedback, going beyond the performance of individual models. We design the structure of the entire system and the role of each component from an ML perspective to enable fast and reliable experimentation and continuous improvement. We align research themes and solution directions with other ML Research Scientists, guiding each research effort towards a unified technical strategy and product outcome. We collaborate with Machine Learning Software Engineers, Data Engineers, Product Teams, etc., to make technical decisions that lead to research outcomes being applied to real products. Requirements PhD degree in Machine Learning or related fields, or equivalent research experience. 5+ years of experience in developing and improving ML-based products or services in the industry. Experience in defining clear ML problems from ambiguous business problems and developing solution strategies based on theoretical rationale and experimentation. Experience in independently and end-to-end solving difficult and challenging problems, and connecting model performance to improvements in user experience or business metrics. Ability to align technical directions and make decisions persuasively through collaboration with other researchers, engineers, and non-technical stakeholders. Preferred Qualifications Experience publishing papers as a primary author in top AI/ML conferences (NeurIPS, ICLR, ICML, CVPR, KDD, etc.). Research experience in key technologies such as recommendation systems, reinforcement learning, LLMs, and AI Agents. Experience in bringing ML systems to market and iteratively improving them for user-facing, high-traffic services. Ability to lead technical discussions with global teams in English. Recruiting Process Employment Type: Full-time Recruiting Process: Resume Screening > Recruiter Call > 1st Interview > 2nd Interview > Final Offer (*Stages may be added/modified as needed). Application Documents: Resume in free form detailing work experience in Korean or English (PDF).
If false information is discovered or there are obstacles to providing employment services in accordance with current laws, the employment contract may be terminated, and the recruitment process may include additional stages and document verifications not previously stated. Individuals eligible for benefits under the Act on the Honorable Treatment of Persons of National Merit (국가보훈대상자) receive preferential treatment in accordance with relevant laws. Please indicate this when applying if you fall into this category, and provide supporting documents upon hiring.
We compare skills, grade, work format and salary
Most often 77,600 to 130,000