SPEAR Lab: Software Performance, Analysis, and Reliability, Concordia University
Tse-Hsun (Peter) Chen

Tse-Hsun (Peter) Chen

Associate Professor, Concordia University Director, Software Performance, Analysis, and Reliability (SPEAR) Lab Academic Associate Member, Mila – Quebec AI Institute

peterc@encs.concordia.ca Google Scholar LinkedIn Twitter

Open to collaboration. I welcome research and industry collaborations; please reach out by email.

Prospective students. I am recruiting PhD and MSc students in coding agents, LLM agents, and software engineering. Former lab members now hold faculty positions in Canada, the US, and India. See how to apply.

I direct the SPEAR lab in the Department of Computer Science and Software Engineering at Concordia, whose software engineering group ranked fifth in the world on CSRankings (2023). We work on making software systems, and the AI agents that increasingly build and operate them, reliable. The lab's main focus today is coding and long-horizon LLM agents, and the benchmarks used to evaluate them.

Our research appears at leading software engineering venues, including ICSE, FSE, ASE, IEEE TSE, and ACM TOSEM, as well as AI venues such as ICML, ACL, and AAAI. Beyond publications, our work has been adopted in practice at companies including BlackBerry, Ericsson, Microsoft, Red Hat, and ERA Environmental. My research is also shaped by industry experience: before joining academia, I worked as a software performance engineer at BlackBerry. Our research contributions have been recognized with the 2024 CS-Can|Info-Can Outstanding Early Career Computer Science Researcher Award.

Research

Reliable coding and long-horizon agents

LLM agents now fix bugs, repair CI failures, and carry out tasks that span hundreds of steps, but they commit to wrong assumptions too early, lose track of what they have done, and fail in ways that are hard to diagnose. We build agents that check each step against execution evidence, keep an accurate record of their state, learn from past runs, and can explain where a run went wrong.

Applications

  • Repository-level program repair and CI repair
  • Reusing past repairs and independently verifying agent-generated patches
  • Multi-step task automation on mobile devices
  • Tracing and attributing failures in agent runs

Representative work: ECLoop, Ledger, STAIR, and RETRACE (2026), ConRAD (ICML '26), Agent-SAMA (AAAI '26), FALAT (2026)

Benchmarks for AI coding agents

Progress on coding agents is only as meaningful as the benchmarks used to measure it. We build repository-level benchmarks grounded in real development workflows, such as CI pipelines, refactoring, and performance work, and study where existing benchmarks overstate what agents can do.

What we evaluate

  • Patch validation through real CI workflows
  • Real-world, repository-level code refactoring
  • Code performance improvements by LLMs

Representative work: CI-Repair-Bench, SWE-Refactor, code performance benchmarks (2026), Probe to Generate (ASE '26)

We also continue our long-standing work on the reliability of large production systems.

Logs and AIOps

Log analysis and operations support that are accurate and efficient enough for production, including LLM methods that run on open-source models on premise.

Used for incident diagnosis, Kubernetes operations with Red Hat OpenShift, and anomaly validation in industrial monitoring.

LibreLog (ICSE '25), LLM agents for AIOps (FSE '26 Industry), Explainable anomaly detection (2026)

Software performance

Detecting performance problems the way users experience them, and focusing performance testing on real usage.

Used for catching slowdowns in mobile OS releases, detecting GUI lag, and load testing.

MobileUPReg (ASE '25), GUIWatcher (ICSE-SEIP '25), MLOLET (ASE '24)

Testing and debugging

Locating faulty code automatically, now with LLM agents that navigate the codebase and reflect on their findings.

Used for shortening debugging time in CI pipelines.

Order Matters (ICSE '26), DepGraph (FSE '24), SBEST (EMSE '25)

Selected publications

Full list on Google Scholar.

News

Lab

Current members

  • Postdoc: Md Ahasanuzzaman
  • PhD: Victor Guerra Veloso, Zeyang Ma, Yi Wen Heng, Md Nakhla Rafi, Linqiang Guo, S M Farah Al Fahim, Yisen Xu, Chenglin Li, Yifei Zhang
  • MSc: Minh Le, Rabeya Khatun Muna, Hassan Jabri

Joining the lab

I am looking for PhD and MSc students interested in LLM agents and software engineering.

Concordia's software engineering group ranked first in Canada, third in North America, and fifth in the world on CSRankings (2023). Through Mila, students in the lab are also connected to Montreal's AI research community. Lab members have received NSERC and FRQNT doctoral scholarships and an ACM SIGSOFT Student Research Competition gold medal.

To apply, please email me your CV, transcripts, and a short note on your research interests.

Alumni

PhD and postdoc

MSc

  • Wasique Islam Shafin 2026
  • Feng Lin 2025
  • Ehsan Abdollahi 2025
  • Lorena Barreto 2024, Slalom Build
  • Misheelt (Mia) Munkhjargal 2024
  • Md Nakhla Rafi 2023, then PhD
  • Tarek Makkouk 2023, Google
  • Zeyang Ma 2022, then PhD
  • Yi Wen Heng 2022, then PhD
  • Steven Locke 2021, Morgan Stanley
  • Zehao Wang 2021, then PhD
  • Dong Jae Kim 2020, then PhD
  • Zi Peng 2020, Tangerine
  • Zhenhao Li 2019, then PhD
  • An Ran Chen 2019, then PhD

Background

Education
PhD and MSc in Computer Science, Queen's University (advisor: Ahmed E. Hassan); BSc in Computer Science, University of British Columbia
Industry
Software Performance Engineer, BlackBerry (2013–2016); Research Consultant, Ericsson AI/ML Upskill Program (2021–2022)
Recognition
CS-Can|Info-Can Outstanding Early Career Computer Science Researcher Award (2024), one of three recipients in Canada; Concordia University Research Award (2022); Gina Cody Research Award (2022); Best New Idea and Emerging Results Paper Award, ICSME (2021); Distinguished Reviewer Awards, ICSE 2026 and ASE 2025
Appointments
PhD Program Director, Department of Computer Science and Software Engineering, Concordia (2025–present); Academic Associate Member, Mila (2026–present)
Service
General Chair, AIware 2026; Reviewer Board, Communications of the ACM; program committees of ICSE, FSE, ASE, NeurIPS, AAAI, ICSME, and MSR; guest editor of special issues in EMSE and JISE
Teaching
SOEN 7481 Software Verification and Testing; SOEN 341 Software Process; SOEN 490 Capstone Design Project