Large Model Algorithm Engineer - Global Payment - Singapore

The Global Payment team of Byte Dance provides payment solutions - including payment acquisitions, disbursements, transaction monitoring, payment method management, foreign exchange conversion, accounting, reconciliations, and so on to ensure that our users have a smooth and secure payment experience on Byte Dance platforms.

The Data Intelligence team in Global Payment leads the efforts of discovering data-driven business values for Global Payment, by collecting, visualising and mining generated data from the pipeline of whole Global Payment system. The outcomes include optimising the key business metrics, improving Global Payment products and user experiences, risk management through data intelligence, predicting the trends from historical data, etc. The LLM Algorithms Team of DI team is responsible for utilizing cutting-edge language and multimodal large model technologies to solve core issues such as user experience and operational efficiency, supporting the implementation of global international payment businesses including risk control, compliance, and credit.

Responsibilities:

• Gain a deep understanding of business requirements, utilize various post-training technologies such as SFT, RLVR, and Agent RL to optimize the performance of models for information extraction, intent classification, tool calling, etc., in scenarios such as intelligent debt collection, intelligent customer service, AIOps, thereby raising the ceiling of business results.
• Based on the understanding of the large model system framework and the boundaries of model capabilities, design and build Agent architectures, including Re Act, Plan-Act, Generate-Verifier, Multi-Agent, etc., and explore Agent Test-Time-Scaling technology.
• Closely follow the latest research results in the LLM field, actively participate in the exploration and research of new businesses, and find the best solutions based on a full understanding of business scenarios.

Requirements:

Minimum Qualifications:

• Possess excellent programming fundamentals and proficiency in at least one programming language such as Python/C++;
• Be familiar with technologies related to NLP, CV, ML, etc., and deep dive into the LLM tech stack (such as Reward Model, GRPO/PPO/DPO, SFT/RFT, etc.);
• Proficiently master common open-source model fine-tuning and training frameworks such as LLama Factory and Verl, be familiar with common inference acceleration frameworks such as vLLM, and understand the principles and implementations of open-source large models such as Qwen and Deep Seek;
• Have excellent ability to analyze and solve problems, be passionate about solving challenging problems, and possess good communication and teamwork skills.

Preferred Qualifications:

• Those with experience in training, fine-tuning, and delivery of well-known large models in the industry are preferred;
• Those with experience in implementing well-known large model application products in the industry or having published papers in top conferences related to large models are preferred;
• Winners of competitions such as ACM/ICPC, NOI/IOI, Top Coder, and Kaggle are preferred.