Research Scientist - Model Capability Boundary Exploration and AI Data Flywheel System Developm

Responsibilities

We are looking for talented individuals to join our team in 2027. As a graduate, you will get opportunities to pursue bold ideas, tackle complex challenges, and unlock limitless growth. Launch your career where inspiration is infinite at our Company.

Successful candidates must be able to commit to an onboarding date by end of year 2027. Please state your availability and graduation date clearly in your resume.

The Applied Machine Learning Enterprise team combines system engineering and machine learning to develop and operate large language model service platforms that offer businesses Model‑as‑a‑Service (MaaS) solutions. The US team drives the design, development, and operation of MaaS solutions across the US and international markets outside mainland China. We are building full‑stack, end‑to‑end solutions spanning text and multimodal LLM algorithms, LLM training/fine‑tuning/inference frameworks, prompt engineering, model alignment, and intelligent agent systems.

Beyond model serving, we operate large‑scale log analytics pipelines that process massive volumes of invocation logs from text models, multimodal models, and agent systems—extracting usage patterns, quality signals, and actionable insights to inform model improvement, system optimization, and product decisions through continuous, data‑driven feedback loops.

With foundation models gradually being applied in real ToB scenarios, AI system optimization now extends beyond the foundation model itself to include a complex business system composed of the model, prompt, memory, tools, skills, workflow, and the external environment. Compared to offline benchmarks, real‑world cases offer greater potential for optimization but also pose challenges such as larger data volumes, higher noise levels, more diverse scenarios, greater structural heterogeneity, and limited user feedback.

Relying on the real‑world data accumulated on the Volcano Ark case platform, this project aims to unify logs, cases, feedback, and environmental information into structured objects that are understandable, attributable, and optimizable. By integrating AI‑assisted tools to guide users in providing efficient feedback, it aims to build an AI data flywheel system tailored to real scenarios. This system will support foundation model iteration and address issues related to environment, memory, tools, and workflows within the business system, focusing on developing agent optimization capabilities that enhance SA/FDE’s efficiency in supporting customers.

• Build a next‑generation big model as a service platform to serve hundreds of LLM‑based applications.
• Develop and maintain the big model as a service platform, including offline training/fine tuning, online inference, model management, and resource orchestration.
• Manage a large number of GPU resources and provide computing power efficiently.

• Currently pursuing or recently completed a Ph.D. in Computer Science, Machine Learning, Artificial Intelligence, Data Science, or a related technical field.
• Research experience in LLM post‑training and alignment, model evaluation, test‑time scaling, agent systems, or large‑scale data curation and optimization.
• Demonstrated research ability through publications, substantial research projects, or internships.
• Ability to work independently on open‑ended research problems from problem formulation to experimental execution.

• Strong interest in foundation models and data‑centric AI, particularly how large models can improve over time through better data, feedback, and system design (data flywheels, continual learning, data curation and valuation, co‑design of algorithms and infrastructure).
• Strong publication record with multiple first‑author papers at top‑tier venues in AI, machine learning, NLP, data mining, or related fields.
• Internship or research experience at an accredited research lab or technology company, ideally involving scalable ML systems with real‑world deployment, feedback loops, or human‑in‑the‑loop pipelines.
• Strong motivation to connect research with practice and build…