Danial Mohseni Taheri

Danial is currently a senior machine learning engineer at Walmart Labs. He is a member of the Core Search Algo team and works on developing and deploying advance NLP and LLM models to enhance the performance of the search results at Walmart.com. Previously, he was a senior machine learning engineer at J.P. Morgan working on advanced NLP algorithms. His efforts was focused on developing AI applications to extract knowledge and information from large scale financial text data.

Before joining J.P. Morgan, he was a Ph.D. candidate at the University of Illinois at Chicago. His research was focused on developing (Deep) machine learning algorithms for problems, including recommendation systems, user behavior modeling, and data-driven decision-making. He was the recipient of multiple international awards, including the best paper award at CEMA 2020-21 and INFORMS 2020. During his PhD, he was focused on:

• Developing (Deep) machine learning algorithms for modeling the users in the applications such as recommendation systems and service system management. He used tools in natural language processing (e.g., Transformers), graph convolution networks, and Bayesian networks to model the time series user data.

• Developing reinforcement learning (RL) algorithms for modeling real systems.

RECENT NEWS:

• November 2019: Our paper "Interpretable User Models via Decision-rule Gaussian Processes" got accepted in Advances in Approximate Bayesian Inference (NeurIPS 2019).

PUBLICATIONS

• Template-aware Attention Model for Earnings Call Report Generation, NewSum EMNLP Workshop 2021.
  
  
  
  
• KATRec: Knowelge-Aware aTtentive Sequential Recommendation System, (with M. Amjadi, T. Tulabandhula)
  
  
  
  
• Interpretable User Models via Decision-rule Gaussian Processes, NeurIPS 2019 (AABI symposium) (with S. Nadarajah, T. Tulabandhula)
  
  Models of user behavior are critical inputs in many prescriptive settings and can be viewed as decision rules that transform state information available to the user into actions. Gaussian processes (GPs), as well as nonlinear extensions thereof, provide a flexible framework to learn user models in conjunction with approximate Bayesian inference. However, the resulting models may not be interpretable in general. We propose decision-rule GPs (DRGPs) that apply GPs in a transformed space defined by decision rules that have immediate interpretability to practitioners. We illustrate this modeling tool on a real application and show that structural variational inference techniques can be used with DRGPs. We find that DRGPs outperform the direct use of GPs in terms of both out-of-sample performance and the quality of optimized decisions. These performance advantages continue to hold when DRGPs are combined with transfer learning.
  
  
• Meeting Corporate Renewbale Power Targets, Under review in Management Science (with S. Nadarajah, A. Trivella) [PDF]
  
  In this paper, we design procurement portfolios for companies which have committed to procuring a percentage of future power demand from renewable sources. Computing multi-stage procurement portfolios for these companies results in intractable Markov decision processes. Using methodologies in approximate dynamic programming, we develop a novel and near-optimal heuristic for decision making on realistic instances.
  
  
• Investment under Limited Long-Term Information, Working Paper (with S. Nadarajah, A. Kelevin, S-E. Feleten )
  
  The limited availability of long-term information about the uncertainties for decision-making problems (i.e., investment in Hydropower plant capacity) increases the potential impact of model misspecification and often lead to suboptimal decisions. To overcome this issue, we propose an algorithm that leverages statistical information in the medium-term and handles the lack of long-term information using robust optimization.
  
  
• Overbooking in Network of Storage Assets, Technical report (with S. Nadarajah, T. Tulabandhula )

EXPERIENCE

SKILLS AND BACKGROUND

INVITED TALKS