Brian Hentschel

I am a Research Scientist at Pinecone focused on the interaction between Information Retrieval and Large Language Models. In particular, I've spent time working on natural language relevance models, such as embedding and reranking models, and on systems that combine retrieval with generation. My work can be most easily seen in Pinecone Assistants.

Before my time at Pinecone, I worked on problems at the intersection of data systems and machine learning. This work was in two directions. The first direction focused on how learning contextual properties of the workload, i.e. expected data and operations, can be used to create new designs that bring significant improvements on fundamental problems in data structures and algorithms. The second direction focused on the opposite direction, building algorithms and systems which make it easier to get insight from data.

Before my Ph.D., I received my undergraduate degree from Pomona College with concentrations in mathematics and computer science, and spent time during my PhD and before as a researcher and software developer in industry with Microsoft Research, IBM Research, LinkedIn, and Amazon.

Please see my resume for more details.


Selected Publications

Exact PPS Sampling with Bounded Sample Size
Brian Hentschel , Peter Haas, and Yuanyuan Tian
Information Processing Letters , 2023.

Entropy-Learned Hashing.
Brian Hentschel , Utku Sirin, and Stratos Idreos.
ACM International Conference on Management of Data (SIGMOD), Philadelphia, Pennsylvania, 2022.

Stacked Filters: Learning to Filter by Structure
Kyle Deeds*,Brian Hentschel*, Stratos Idreos.
International Conference on Very Large Databases (VLDB), 2021.
[* denotes co-first authors].

MotherNets: Rapid Deep Ensemble Learning.
Abdul Wasay, Brian Hentschel, Yuze Liao, Sanyuan Chen, and Stratos Idreos.
Conference on Machine Learning and Systems (MLSys), 2020.

General Temporally-Biased Sampling for Online Model Management.
Brian Hentschel , Peter J. Haas, Yuanyuan Tian.
Transactions on Database Systems (TODS), 2019.

Learning Data Structure Alchemy.
Stratos Idreos, Kostas Zoumpatianos, Subarna Chatterjee, Wilson Qin, Abdul Wasay, Brian Hentschel, Mike Kester, Niv Dayan, Demi Guo, Minseo Kang, and Yiyou Sun
Bulletin of the IEEE Computer Society Technical Committee(Data Engineering), 2019.

Temporally Biased Sampling for Online Model Management.
Brian Hentschel, Peter Haas, and Yuanyuan Tian.
International Conference on Extending Database Technology (EDBT), 2018.
[Best Paper, ACM SIGMOD Research Highlight]

The Periodic Table of Data Structures.
Stratos Idreos, Konstantinos Zoumpatianos, Manos Athanassoulis, Niv Dayan, Brian Hentschel, Michael S. Kester, Demi Guo, Lukas Maas, Wilson Qin, Abdul Wasay, and Yiyou Sun.
Bulletin of the IEEE Computer Society Technical Committee (Data Engineering), 2018.

Column Sketches: A Scan Accelerator for Rapid and Robust Predicate Evaluation
Brian Hentschel, Michael Kester, and Stratos Idreos.
ACM International Conference on Management of Data (SIGMOD), 2018.

The Data Calculator: Data Structure Design and Cost Synthesis From First Principles, and Learned Cost Models.
Stratos Idreos, Kostas Zoumpatianos, Brian Hentschel, Michael Kester, and Demi Guo.
ACM International Conference on Management of Data (SIGMOD), 2018.

Teaching

I have thoroughly enjoyed my time teaching at both Harvard University and in my undergrad at Pomona College. Below are a list of courses I have helped with.

Harvard University
Teaching Fellow at CS165: Data Systems | 2015-2019,2021
Harvard Distinction in Teaching 2016, 2019

Harvard University
Teaching Fellow at CS265: Advanced Data Systems | 2016, 2020

Pomona College
Computer Science Mentor at CS 81: Computability and Logic | Spring, Fall 2012
Peer Tutor at Math 131: Real Analysis | Fall 2013

Contact

Email: bhentschel9 [at] gmail.com