Prospective students/interns, read this first!

I'm Laurent Bindschaedler

I'm a Researcher working on Systems for Big Data, Graph Processing, and Machine Learning

Affiliation Max Planck Institute for Software Systems
Address Room 409, MPI-SWS, Campus E 1 5, 66123 Saarbrücken, Germany
Contact bindsch AT mpi-sws DOT org
Hiring Students starting in 2025

I am a Research Group Leader at the Max Planck Institute for Software Systems (MPI-SWS) where I lead the Data Systems Group (DSG). I build systems for Big Data, Graph Processing, and Machine Learning. I am also interested in improving Systems with Machine Learning. Prior to joining MPI-SWS, I was a Postdoctoral Fellow at MIT CSAIL working with Prof. Tim Kraska. In 2020, I completed my Ph.D. at EPFL, Switzerland, advised by Prof. Willy Zwaenepoel. Before my graduate studies, I worked on my own Cloud Computing startup.

Research Interests

Big Data

Distributed Systems

Software Architecture

Machine Learning

Cloud Computing

Security & Privacy

Publications

Towards Reliable Latent Knowledge Estimation in LLMs: Zero-Prompt Many-Shot Based Factual Knowledge Extraction

In 18th ACM International Conference on Web Search and Data Mining (2025).
Qinyuan Wu, Mohammad Aflah Khan, Soumi Das, Vedant Nanda, Bishwamittra Ghosh, Camila Kolling, Till Speicher, Laurent Bindschaedler, Krishna Gummadi, and Evimaria Terzi.
[ PDF] [ BibTeX]

In this paper, we focus on the challenging task of reliably estimating factual knowledge that is embedded inside large language models (LLMs). To avoid reliability concerns with prior approaches, we propose to eliminate prompt engineering when probing LLMs for factual knowledge. Our approach, called Zero-Prompt Latent Knowledge Estimator (ZP-LKE), leverages the in-context learning ability of LLMs to communicate both the factual knowledge question as well as the expected answer format. Our knowledge estimator is both conceptually simpler (i.e., doesn’t depend on meta-linguistic judgments of LLMs) and easier to apply (i.e., is not LLM-specific), and we demonstrate that it can surface more of the latent knowledge embedded in LLMs. We also investigate how different design choices affect the performance of ZP-LKE. Using the proposed estimator, we perform a large-scale evaluation of the factual knowledge of a variety of open-source LLMs, like OPT, Pythia, Llama(2), Mistral, Gemma, etc. over a large set of relations and facts from the Wikidata knowledge base. We observe differences in the factual knowledge between different model families and models of different sizes, that some relations are consistently better known than others but that models differ in the precise facts they know, and differences in the knowledge of base models and their finetuned counterparts.

Show Abstract

Hide Abstract

AlterEgo: A Dedicated Blockchain Node For Analytics

In 7th International Workshop on Edge Systems, Analytics and Networking (2024).
Qi Guo, Mahdi Alizadeh, Ali Falahati, and Laurent Bindschaedler.
[ PDF] [ BibTeX]

Blockchains today amass terabytes of transaction data that demand efficient and insightful real-time analytics for applications such as smart contract hack detection, price arbitrage on decentralized exchanges, or trending token analysis. Conventional blockchain nodes, constrained by their RPC APIs, and specialized ETL-based blockchain analytics systems grapple with a trade-off between materializing pre-calculated query results and analytical expressiveness. In response, we introduce AlterEgo, a blockchain node architected specifically for analytics that maintains parity with traditional nodes in ingesting consensus-produced blocks while integrating a robust analytics API. Our prototype supports efficient transactional and analytical processing while circumventing the rigidity of ETL workflows, offering a better trust model, enabling distributed and collaborative querying, and achieving significant performance improvements over the state-of-the-art.

Show Abstract

Hide Abstract

Unshackling Database Benchmarking from Synthetic Workloads

In Demo of the 39th IEEE International Conference on Data Engineering (2023).
Parimarjan Negi*, Laurent Bindschaedler*, Mohammad Alizadeh, Tim Kraska, Jyoti Leeka, Anja Gruenheid, and Matteo Interlandi.
[ PDF] [ BibTeX]

Introducing new (learned) features into a DBMS requires considerable experimentation and benchmarking to avoid regressions in production (customer) workloads. Using standard benchmarks such as TPC-H and TCH-DS is common practice, but, unfortunately, these do not represent the complexity of real production workloads. To solve this problem, in this demo, we propose a technique that generates a synthetic dataset from query logs and metadata—without touching the original data. The keystone of our approach is to map the data generation as a SAT problem where constraints, such as runtime cardinalities, are extracted from query logs and metadata. We show that our approach can generate representative benchmarks mirroring the performance of the original data without trading off privacy. The demo will guide the attendees through the various steps involved in the data generation and testing process.

Show Abstract

Hide Abstract

Accurate Automatic Camera Calibration on Low-Quality CCTV Traffic Video Streams

In Proceedings of the International Conference on Computers and Automation (2022).
Audhav Durai, Sankarshanaram Vempati, and Laurent Bindschaedler.
[ PDF] [ BibTeX]

This paper focuses on extracting vehicle trajectory from low-quality, uncalibrated traffic cameras via fully automatic camera calibration and homography transformations. We analyze video streams from 511 traffic surveillance cameras in varying areas and arbitrary relative locations and orientations to roads. We propose a novel automatic calibration methodology that is used to transform the video streams to a top-down perspective from a minute of the video stream. Using this perspective, we can extract both lane changes and vehicle speed. Our automatic camera calibration algorithm is inspired by two vanishing point based automatic camera calibration methods. We implement more accurate and robust methods for the detection of both vanishing points. We show that YOLOv4 combined with DeepSORT is 12.5 times faster than the leading vehicle detection model and achieves a 7.81% higher average precision and 6.04% higher mean average precision in low quality traffic surveillance cameras. We use this model for a more robust estimation of the first vanishing point and more robust object tracking. Furthermore, we propose a fast, novel “guess and check” algorithm for the detection of the second vanishing point to ensure accurate second vanishing point detection. We show that our camera calibration algorithm produces fewer inaccuracies than the state-of-the-art automatic camera calibration methodology in through qualitative results.

Show Abstract

Hide Abstract

Tesseract: Distributed, General Graph Pattern Mining on Evolving Graphs

In Proceedings of the 16th European Conference on Computer Systems (2021).
Laurent Bindschaedler, Jasmina Malicevic, Baptiste Lepers, Ashvin Goel, and Willy Zwaenepoel.
[ PDF] [ Slides] [ Video] [ Short Video] [ Code] [ BibTeX]

Tesseract is the first distributed system for executing general graph mining algorithms on evolving graphs. Tesseract scales out by decomposing a stream of graph updates into per-update mining tasks and dynamically assigning these tasks to a set of distributed workers. We support this design using a novel approach to change detection that efficiently determines the exact modifications to the algorithm’s output for each update to the input graph. We use a disaggregated, multiversioned graph store to allow workers to process updates independently, without producing duplicates. Moreover, Tesseract provides interactive mining insights for complex applications using our incremental aggregation API. Finally, we implement and evaluate Tesseract and demonstrate that it achieves orders-of-magnitude improvements over state-of-the-art systems.

Show Abstract

Hide Abstract

Towards a Benchmark for Learned Systems

In Proceedings of the 12th International Workshop on Self-Managing Database Systems (2021).
Laurent Bindschaedler, Andreas Kipf, Tim Kraska, Ryan Marcus, and Umar Farooq Minhas.
[ PDF] [ Slides] [ Video] [ BibTeX]

This paper aims to initiate a discussion around benchmarking data management systems with machine-learned components. Traditional benchmarks such as TPC or YCSB are insufficient to analyze and understand these learned systems because they evaluate the performance under a stable workload and data distribution. Learned systems automatically specialize and adapt database components to a changing workload, database, and execution environment, thereby making conventional metrics such as average throughput ill-suited to understand their performance fully. Moreover, the standard cost-per-performance metrics fail to account for essential trade-offs related to the training cost of models and the elimination of manual database tuning. We present several ideas for designing new benchmarks that are better suited to evaluate learned systems. The main challenges entail developing new metrics to capture the particularities of learned systems and ensuring that benchmark results remain comparable across many deployments with wide-ranging designs.

Show Abstract

Hide Abstract

Hailstorm: Disaggregated Compute and Storage for Distributed LSM-based Databases

In Proceedings of the 25th International Conference on Architectural Support for Programming Languages and Operating Systems (2020).
Laurent Bindschaedler, Ashvin Goel, and Willy Zwaenepoel.
[ PDF] [ Slides] [ Video] [ Code] [ BibTeX]

Distributed LSM-based databases face throughput and latency issues due to load imbalance across instances and interference from background tasks such as flushing, compaction, and data migration. Hailstorm addresses these problems by deploying the database storage engines over a distributed filesystem that disaggregates storage from processing, enabling storage pooling and compaction offloading. Hailstorm pools storage devices within a rack, allowing each storage engine to fully utilize the aggregate rack storage capacity and bandwidth. Storage pooling successfully handles load imbalance without the need for resharding. Hailstorm offloads compaction tasks to remote nodes, distributing their impact, and improving overall system throughput and response time. We show that Hailstorm achieves load balance in many MongoDB deployments with skewed workloads, improving the average throughput by 60%, while decreasing tail latency by as much as 5X. In workloads with range queries, Hailstorm provides up to 22X throughput improvements. Hailstorm also enables cost savings of 47-56% in OLTP workloads.

Show Abstract

Hide Abstract

Rock You Like a Hurricane: Taming Skew in Large Scale Analytics

In Proceedings of the 13th EuroSys Conference (2018).
Laurent Bindschaedler, Jasmina Malicevic, Nicolas Schiper, Ashvin Goel, and Willy Zwaenepoel.
[ PDF] [ Slides] [ Video] [ Code] [ BibTeX]

Current cluster computing frameworks suffer from load imbalance and limited parallelism due to skewed data distributions, processing times, and machine speeds. We observe that the underlying cause for these issues in current systems is that they partition work statically. Hurricane is a high-performance large-scale data analytics system that successfully tames skew in novel ways. Hurricane performs adaptive work partitioning based on load observed by nodes at runtime. Overloaded nodes can spawn clones of their tasks at any point during their execution, with each clone processing a subset of the original data. This allows the system to adapt to load imbalance and dynamically adjust task parallelism to gracefully handle skew. We support this design by spreading data across all nodes and allowing nodes to retrieve data in a decentralized way. The result is that Hurricane automatically balances load across tasks, ensuring fast completion times. We evaluate Hurricane’s performance on typical analytics workloads and show that it significantly outperforms state-of-the-art systems for both uniform and skewed datasets, because it ensures good CPU and storage utilization in all cases.

Show Abstract

Hide Abstract

Chaos: Scale-out Graph Processing from Secondary Storage

In Proceedings of the 25th Symposium on Operating Systems Principles (2015).
Amitabha Roy, Laurent Bindschaedler, Jasmina Malicevic, and Willy Zwaenepoel.
[ PDF] [ Slides] [ Video] [ Code] [ BibTeX]

Chaos scales graph processing from secondary storage to multiple machines in a cluster. Earlier systems that process graphs from secondary storage are restricted to a single machine, and therefore limited by the bandwidth and capacity of the storage system on a single machine. Chaos is limited only by the aggregate bandwidth and capacity of all storage devices in the entire cluster. Chaos builds on the streaming partitions introduced by X-Stream in order to achieve sequential access to storage, but parallelizes the execution of streaming partitions. Chaos is novel in three ways. First, Chaos partitions for sequential storage access, rather than for locality and load balance, resulting in much lower pre-processing times. Second, Chaos distributes graph data uniformly randomly across the cluster and does not attempt to achieve locality, based on the observation that in a small cluster network bandwidth far outstrips storage bandwidth. Third, Chaos uses work stealing to allow multiple machines to work on a single partition, thereby achieving load balance at runtime. In terms of performance scaling, on 32 machines Chaos takes on average only 1.66 times longer to process a graph 32 times larger than on a single machine. In terms of capacity scaling, Chaos is capable of handling a graph with 1 trillion edges representing 16 TB of input data, a new milestone for graph processing capacity on a small commodity cluster.

Show Abstract

Hide Abstract

Benchmarking X-Stream and Graphchi

EPFL Technical Report (2014).
Laurent Bindschaedler, and Amitabha Roy.
[ PDF] [ Code] [ BibTeX]

A number of graph processing systems have been proposed that operates using secondary storage on a single machine. In the interests of furthering research we (LABOS, EPFL) and others (Graphchi, CMU) have released the source code of X-Stream and Graphchi as examples of such systems in source code form. The initial set of comparisons between X-Stream and Graphchi in the SOSP 2013 paper were limited to a one type of SSD. We extend those benchmarks in this short report by considering a variety of storage devices. In addition, our own experience with benchmarking both these systems has shown that it can often be tricky for newcomers to property set up and benchmark them. Therefore we present detailed notes on our benchmark runs to ensure that others can replicate them and avoid subtle errors.

Show Abstract

Hide Abstract

Big Brother Knows Your Friends: on Privacy of Social Communities in Pervasive Networks

In Proceedings of the 10th International Conference on Pervasive Computing (2012).
Igor Bilogrevic, Murtuza Jadliwala, Istvan Lam, Imad Aad, Philipp Ginzboorg, Valtteri Niemi, Laurent Bindschaedler, and Jean-Pierre Hubaux.
[ PDF] [ BibTeX]

Location privacy is a major concern in an increasingly connected and highly pervasive network of mobile users. Novel location-based applications and device-to-device services (on these mobile devices) are gaining popularity, but at the same time, these services allow curious service providers and eavesdroppers to track users and their movements. Earlier research efforts on location-privacy preservation, which were mostly based on identifier-change mechanisms in spatio-temporal de-correlation regions called mix-zones, show that coordinated identifier-change techniques are reasonably effective in a simulation setting, although some smart attacks are still possible. However, a thorough analysis of these mechanisms that takes into consideration communication patterns and mobility from a real-life deployment is missing from these results. In this paper, we evaluate in a real-life setting the effectiveness of standard mix-zone-based privacy protection mechanisms against probabilistic tracking attacks. Our exper- iments involved 80 volunteers carrying smartphones for 4 months and being constantly eavesdropped on an adversarial mesh network of standard wireless Access Points (APs). To the best of our knowledge, this is the first study that provides empirical evidence about the effectiveness of mix-zone-based privacy-preserving mechanisms against practical adversaries in upcoming wireless and mobile systems.

Show Abstract

Hide Abstract

Track Me If You Can: on the Effectiveness of Context-based Identifier Changes in Deployed Mobile Networks

In Proceedings of the 19th Annual Network & Distributed System Security Symposium (2012).
Laurent Bindschaedler, Murtuza Jadliwala, Igor Bilogrevic, Imad Aad, Philipp Ginzboorg, Valtteri Niemi, and Jean-Pierre Hubaux.
[ PDF] [ Slides] [ Code] [ BibTeX]

Show Abstract

Hide Abstract

Making Mobile Augmented Reality A Reality

In Proceedings of the 13th International Conference on Human Computer Interaction with Mobile Devices and Services (2011).
Laurent Bindschaedler, Hendrik Knoche, and Jeffrey Huang.
[ PDF] [ Code] [ BibTeX]

Recent advances in mobile device technology have freed augmented reality (AR) applications from the constraints of desktops, laptops and head-mounted displays. But they are met with a lack of guidelines on the design and user interactions of mobile device-based AR systems. The situation for developers is further exacerbated by closed-license environments and inflexible solutions. We provide an overview of the design of AR applications on handheld devices, the necessary building blocks and problems that future AR systems need to overcome. This experience was gathered during the design and development of an AR framework for the Android™ platform. User experience evaluations showed a great demand for overlay collision avoidance and the value of being able to freeze AR screens. These will be valuable for the design of future mobile device based AR applications.

Show Abstract

Hide Abstract

Portfolio

Hailstorm

Software

Hurricane

Software

Chaos

Software

Work Experience

2022 -

Research Group Leader

Max Planck Institute for Software Systems

2020 - 2022

Postdoctoral Fellow

Data Systems Group, CSAIL

Host: Prof. Tim Kraska

2015 - 2020

PhD Candidate

LABOS, IC Faculty

Thesis Advisor: Prof. Willy Zwaenepoel

2014 - 2015

Scientific Assistant

LABOS, IC Faculty

2013-2014

Consultant & Head of IT

2011 - 2013

Co-Founder & VP of Engineering

2011 - 2011

Research Intern

2010 - 2011

Student Assistant

IC Faculty

2006 - 2018

IT Pioneer

Staff 1st Region — Swiss Armed Forces

Education

2015 - 2020

PhD in Computer Science

2009 - 2011

Master in Computer Science

2006 - 2009

Bachelor in Computer Science

Awards and Achievements

Early Postdoc.Mobility — 2020

Awarded an Early Postdoc.Mobility Fellowship by the Swiss National Science Foundation. [ Link]

Best Video Award — ASPLOS Conference — 2020

“Hailstorm: Disaggregated Compute and Storage for Distributed LSM-based Databases” was voted best presentation at the ASPLOS’20 conference. [ Link] [ Watch]

Teaching Assistant Award — EPFL IC Faculty — 2017

Awarded by EPFL’s School of Computer and Communication Sciences for excellent contributions to teaching. [ Link]

Public Prize Winner — Exposure Movie Hackathon — 2017

Exposure is a 3-day hackathon that brings together scientists and filmmakers to create short science films. “Jenna & DANBOT”, our movie about Artificial Intelligence and Machine Learning was voted best film by the public. [ Link] [ IMDb] [ Watch]

PhD Fellowship — EDIC EPFL — 2015

Awarded by EPFL’s Computer and Communication Sciences Doctoral Program (EDIC). EDIC awards promising students a one-year fellowship. [ Link]

Round 1 Winner — Venture Kick — 2012

Venture Kick is an initiative to support Swiss startups with enough funding to kick-start their entrepreneurial success. Awarded a first round grant with LakeMind. [ Link]