Financial Foundation LLM Architect

DiDi Chuxing; San Jose, Northern (CA, KY); Hybrid; USD 180,000 - 240,000; Full time

Job summary

DiDi Global Inc. is hiring an Algorithm Expert to build foundation-model based financial risk solutions in a fast‑growing payments and mobility context. You will design pre‑training pipelines, construct cross‑domain sequence representations, and run robust ablations to inform architecture decisions.

The role involves production deployment of risk models, blending representations into downstream systems, and close collaboration with product and engineering teams in San Jose, CA.

Qualifications

• Must have a Master’s degree or higher in Computer Science, Mathematics, Statistics, or related field.
• 3+ years of deep learning algorithm R&D experience and full training pipeline from scratch.
• Proficiency with Transformer architectures (GPT/BERT variants) and practical sequence modeling.
• Hands-on experience with PyTorch and distributed multi-GPU training.
• Strong experimental design skills and ability to perform ablations.

Responsibilities

• Design and implement pre-training pipelines for financial behavior foundation models (objective selection, tokenization, scaling).
• Build unified sequence representations across payments, ride-hailing, delivery, and credit domains.
• Develop ablation studies and evaluation frameworks to guide architecture decisions.
• Integrate pre-trained representations into risk-control downstream tasks and online deployment.

Skills

Foundation models

Transformer architectures

PyTorch

Sequence modeling

Experimental design

Model deployment

Time-series modeling

Graph neural networks

Education

Master's degree or above in CS/Math/Stats

Tools

PyTorch

Distributed training

Time-series tools

Job description

DiDi Global Inc. is hiring an Algorithm Expert to build foundation-model based financial risk solutions in a fast‑growing payments and mobility context. You will design pre‑training pipelines, construct cross‑domain sequence representations, and run robust ablations to inform architecture decisions.

The role involves production deployment of risk models, blending representations into downstream systems, and close collaboration with product and engineering teams in San Jose, CA.