The Team
Systems Excellence, Industry Execution
Teyon is led by systems researchers who has spent nearly two decades on deterministic replay, memory analysis, and automatic failure diagnosis, and who has run large-scale ML training in production.
Tongping Liu
CEO & Tech Founder
Pioneer in Deterministic Systems, PhD
- Ex-tenured professor at UMass and principal engineer in AI infrastructure
- 18+ years in systems reliability, published at SOSP and PLDI
- Led large-scale ML training at ByteDance and XPeng, improving training throughput by over 50%
- GPU and CPU memory management and profiling
- 12 US patents granted and 24 provisional patents, 14 owned by Teyon
Research
A decade of systems research, now in production
Teyon's engines come from peer-reviewed work on deterministic replay, memory analysis, and automatic failure diagnosis at SOSP, PLDI, OOPSLA, ICSE, CCS, and USENIX Security.
Papers
8 published · 1 in progressSelected papers; the original PDF available for download in the detail pages.
MemPerf: Profiling Allocator-Induced Performance Slowdowns
The first profiler that tells you when the memory allocator, not your code, is what slows a program down.
Read the summaryNUMAlloc: A Faster NUMA Memory Allocator
A drop-in allocator built for NUMA machines: 15.7% faster than mimalloc and 20.9% faster than glibc, scaling to 128 threads.
Read the summaryWatcher: In-Situ Failure Diagnosis
Pinpoints the root cause of a crash inside the failing process, in seconds, without a memory dump ever leaving the machine.
Read the summaryiReplayer: In-Situ and Identical Record-and-Replay for Multithreaded Applications
Records multithreaded programs at about 3% overhead and replays them identically inside the original process.
Read the summaryGuarder: A Tunable Secure Allocator
A secure allocator with stable, tunable randomization entropy at under 3% overhead.
Read the summaryFreeGuard: A Faster Secure Heap Allocator
A secure heap allocator that defends against overflows, use-after-frees, and double frees at under 2% overhead.
Read the summaryDoubleTake: Fast and Precise Error Detection via Evidence-Based Dynamic Analysis
Finds buffer overflows, use-after-frees, and leaks at under 5% overhead by checking for evidence at epoch ends and replaying only when something went wrong.
Read the summaryDthreads: Efficient Deterministic Multithreading
Makes unmodified C/C++ multithreaded programs deterministic, often at the speed of pthreads.
Read the summaryMemScope
Ongoing work on CPU and GPU memory analysis, the research basis of XScope.
Patents
12 granted · 24 provisionalGranted US patents naming Tongping Liu as inventor, assigned to the universities and companies where the work was done. A further 24 provisional patents are on file, 14 of them owned by Teyon; provisional filings are not public, so only granted patents are listed.
- US 12,541,443Classification of Different Types of Cache MissesUniversity of Massachusetts Amherst · 2026
- US 12,314,761Memory Allocation and Management in Non-Uniform Memory Access (NUMA) Computing EnvironmentsUniversity of Massachusetts Amherst · 2025
- US 11,599,445Watcher: Precise and Fully-Automatic On-Site Failure DiagnosisUniversity of Texas System · 2023
- US 11,593,483Guarder: An Efficient Heap Allocator with Strongest and Tunable SecurityUniversity of Texas System · 2023
- US 11,294,652Low-Overhead Detection Techniques for Synchronization Problems in Parallel and Concurrent SoftwareUniversity of Texas System · 2022
- US 10,915,424Defeating Deadlocks in Production SoftwareUniversity of Texas System · 2021
- US 10,901,828FreeGuard: A Faster Secure Heap AllocatorUniversity of Texas System · 2021
- US 10,474,369Mapping Guest Pages to Disk Blocks to Improve Virtual Machine Management ProcessesVMware · 2019
- US 10,394,714System and Method for False Sharing PredictionFuturewei Technologies · 2019
- US 9,678,883System and Method for Detecting False SharingFuturewei Technologies · 2017
- US 9,547,599System and Method for Predicting False SharingFuturewei Technologies · 2017
- US 8,868,849Coupled Lock Allocation and Lookup for Shared Data Synchronization in Symmetric Multithreading EnvironmentsSamsung Electronics · 2014
Strategic Advisors & Early Backers
Backed by Industry Leaders
Backed by engineering leaders from DeepMind, Intel, and enterprise infrastructure.
Bin Ni
Strategic Advisor
Pioneer in AI Coding
- Former Head of AI Coding, DeepMind
- Led 13 Google moonshot announcements
Todd Blaschka
GTM Advisor
Category Builder in Enterprise Infra
- CRO, TigerGraph; VP of Sales, Clustrix
- 25+ years of B2B infrastructure go-to-market, including PKWare and IBM
Chen Tian
Execution & Angel
Veteran in Hyper-Scale Infrastructure
- VP of Engineering, Inceptio
- Enterprise-grade AI cloud and in-vehicle infrastructure
Youfeng Wu
Technical Advisor
Expert in Binary Analysis
- Former Programming Systems Director, Intel
- 30+ years as an authority in systems performance
Ready for resilient AI infrastructure?
Stop losing $10K+ per GPU failure. Let's talk.