DCDaryl Chang

Daryl Chang

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Daryl Chang has a total of 16 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Self-Evolving Recommendation System: End-To-End Autonomous Model Optimization With LLM Agents

    by Haochen Wang, Yi Wu, Daryl Chang, Li Wei, Lukasz Heldt

    Optimizing large-scale machine learning systems, such as recommendation models for global video platforms, requires navigating a massive hyperparameter search space and, more critically, designing sophisticated optimizers, architectures, and reward functions to capture nuanced user behaviors. Achieving substantial improvements in these areas is a non-trivial task, traditionally relying on extensive manual iterations to test new hypotheses. We propose a self-evolving system that leverages Large Language Models (LLMs), specifically those from Google’s Gemini family, to autonomously generate, train, and deploy high-performing, complex model changes within an end-to-end automated workflow. The self-evolving system consists of an Offline Agent (Fast Loop) that performs high-throughput hypothesis generation to optimize for proxy metrics, and an Online Agent (Slow Loop) that validates candidates against delayed north star business metrics in live production. Our agents act as specialized Machine Learning Engineers (MLEs): they exhibit deep reasoning capabilities, discovering novel improvements in optimization algorithms and model architecture, and formulating innovative reward functions that target long-term user engagement. The effectiveness of this approach is demonstrated through several successful production launches at YouTube, confirming that autonomous, LLM-driven evolution can surpass traditional engineering workflows in both development velocity and model performance.


  • Open Access

    Learned Ranking Function: From Short-term Behavior Predictions to Long-term User Satisfaction

    by Yi Wu, Daryl Chang, Jennifer She, Zhe Zhao, Li Wei, Lukasz Heldt

    We present the Learned Ranking Function (LRF), a system that takes short-term user-item behavior predictions as input and outputs a slate of recommendations that directly optimizes for long-term user satisfaction. Most previous work is based on optimizing the hyperparameters of a heuristic function. We propose to model the problem directly as a slate optimization problem with the objective of maximizing long-term user satisfaction. We also develop a novel constraint optimization algorithm that stabilizes objective tradeoffs for multi-objective optimization. We evaluate our approach with live experiments and describe its deployment on YouTube.


  • Open Access

    Improving Training Stability for Multitask Ranking Models in Recommender Systems

    by Jiaxi Tang, Yoel Drori, Daryl Chang, Maheswaran Sathiamoorthy, Justin Gilmer, Li Wei, Xinyang Yi, Lichan Hong, Ed H. Chi

    Recommender systems play an important role in many content platforms. While most recommendation research is dedicated to designing better models to improve user experience, we found that research on stabilizing the training for such models is severely under-explored. As recommendation models become larger and more sophisticated, they are more susceptible to training instability issues, i.e., loss divergence, which can make the model unusable, waste significant resources and block model developments. In this paper, we share our findings and best practices we learned for improving the training stability of a real-world multitask ranking model for YouTube recommendations. We show some properties of the model that lead to unstable training and conjecture on the causes. Furthermore, based on our observations of training dynamics near the point of training instability, we hypothesize why existing solutions would fail, and propose a new algorithm to mitigate the limitations of existing solutions. Our experiments on YouTube production dataset show the proposed algorithm can significantly improve training stability while not compromising convergence, comparing with several commonly used baseline methods.


  • Pushing Malaysia's Drilling Industry into a New Frontier: How a Distinctive Wellhead Design Enable Implementation of a Fully Offline Well Cementing Resulting in Significant Shift in Operational Efficiency

    by Fauzi Abbas, Azrynizam M. Nor, Daryl Chang, Wojciech Czapor

    Traditionally, rigs are positioned over a well from the moment the surface casing is drilled till the installation of the wellhead Christmas Tree. This resulted in the loss of precious rig time as the rig is idling during online cementing. However, in Field A, offshore of Terengganu, Malaysia, a new approach has managed to eliminate such inefficiency drastically. This paper discusses the employment of offline cementing techniques made possible with distinctive wellhead technology, that enabled a fast, efficient, and safe drilling and completion. In this offshore drilling program, five wells were successfully batched drilled and cemented offline by ingenious planning and skidding of the rig. As the first well has reached its depth, offline adapters with pressure compatible cementing heads and accessories are installed on the wellhead. This allowed the rig to skid to the adjacent wells to continue drilling while the previous well was being cemented with the use of full bore offline cementing adapters. Full bore offline cementing adapters helped maintained familiarity by allowing the use of simple conventional wiper plugs. These offline adapters are compatible with both mandrel and slip hangers and have pressure bleed off and monitoring capabilities. These steps were then repeated in all of the wells, from the surface casing till the tubing of these mono bore completions. Besides that, production and casing hangers come with barrier plug profiles allowing plugs to be pre-made up and tested. This reduced installation time on the rig floor and ensure that there was a positively tested barrier, before nippling down the riser. Hangers with seals also allowed for a single trip installation, further reducing rig time. The five wells were drilled and completed in record-breaking time at an average of 7 days per well and below budget, resulting in a time saving of 31 hours and 18.5% of cumulative savings. In conclusion, this wellhead technology introduced a new approach to drilling and completion allowing vast improvement to the field economics and profitability to the development of marginal fields.