👨💻 Bernard Ladenthin | Homepage | LinkedIn
👋 Hey — nice that you want to know more about me.
🌍 Based near Berlin, Germany.
I'm a Senior Site Reliability Engineer at Deutsche Bank, focused on stabilizing mission-critical financial systems, cloud-native microservices, and production reliability at scale. Different domain, same discipline as everything below.
What I like best is the smallest possible change to the hardest possible problem: a fault buried deep in a large, unfamiliar, sparsely documented system — narrowed down until the fix is a handful of lines, and provable.
The loop rarely changes, and every pass is meant to make the next one cheaper:
- Reproduce it, then make the reproduction cheap enough to run again and again
- Measure instead of argue — a plausible theory that survives an hour of reasoning still dies in one experiment
- Halve the problem until what remains is a minimal, standalone reproducer
- Change exactly one thing, and show the red-to-green transition
- Write down what turned out to be wrong, not just what turned out to be right
The terrain is usually the same, and it is why the method has to be this literal-minded: binary analysis — disassembly, undocumented file and data formats — emulation, and firmware- and boot-level debugging. Systems with neither source nor documentation to read, where the only honest answer comes from measuring the difference between a run that works and a run that does not.
📡 Off the clock — the same interests turn into electronics projects, low-level optimization, and small utilities and microtools that sit between software and hardware. Older and less common architectures and AIX/Unix systems fascinate me, and I'm still learning my way around them.
Java is my primary language. Everything else on this list is picked to reach whichever layer the problem actually lives on — and interesting problems rarely stay on one.
| Layer | What I work with |
|---|---|
| Application / JVM | Java (primary), Kotlin, Groovy — Maven Central libraries, CLI tools, services |
| Between the layers | JNI and native bindings, JNR-FFI, OpenCL from the JVM, memory-mapped storage (LMDB), stream and process plumbing, JVM bytecode |
| Systems / native | C, C++, POSIX, cross-platform portability across Linux, macOS, Windows, the BSDs and AIX |
| Low level | Assembler (x86, PowerPC), disassembly |
Cutting across every layer:
| Area | What I work with |
|---|---|
| Scripting & glue | Python, Bash, PowerShell |
| OS & systems | GNU/Linux, Unix / AIX, Windows, WINE, QEMU/KVM |
| Data & integrity | SQL, LMDB key–value store, error correction (ECC memory, FEC, Reed–Solomon) |
| Performance | high-performance and parallel computing — OpenCL, GPU kernels, lock-free concurrency |
| Cryptography | elliptic curves (ECC), RSA, Diffie–Hellman, blockchain (Bitcoin, Monero, Bitmessage) |
| Build & DevOps | Maven, Gradle, Ant, Jenkins, GitHub Actions, reproducible releases & artifact signing |
| Privacy & security | TOR, secure communication, blockchain privacy, attack-vector analysis |
🧪 Quality & testing — the part I am most opinionated about. Test-driven development, and then everything that can be made to fail a build:
| Kind | What I use |
|---|---|
| Property-based & architecture tests | jqwik, ArchUnit |
| Mutation testing & coverage | PIT (gated at 100%), JaCoCo |
| Static analysis & null safety | SpotBugs, Error Prone, NullAway, Checker Framework, JSpecify |
| Concurrency verification | jcstress, Lincheck, vmlens |
| Project | What it does |
|---|---|
| 🧠 BitcoinAddressFinder | High-performance JVM + OpenCL tool that generates and checks Bitcoin and altcoin addresses at scale, built for cryptographic experiments and raw key-search throughput — parallel EC key generation, vanity-address search, fast LMDB lookups. |
| 🔄 streambuffer | Thread-safe buffer connecting an OutputStream to an InputStream for real-time data flow between threads: concurrent reads and writes, automatic buffer trimming, optional clone-on-write safety. |
| 🦙 java-llama.cpp | Java bindings for llama.cpp that run local LLMs entirely on the JVM — text and chat completion, embeddings, reranking, infilling, no cloud dependency. Pre-built native binaries for Linux, macOS, Windows and Android; on Maven Central as net.ladenthin:llama. |
| 🗂️ srcmorph | Prompt-driven source-tree transformer that documents and reshapes Java projects with a local llama.cpp model, no cloud required. Walks the tree through configurable prompts, emits layered per-file, per-package and per-project .ai.md summaries, and updates only what changed. 3-module Maven reactor: core (net.ladenthin:srcmorph), CLI (net.ladenthin:srcmorph-cli), Maven plugin (net.ladenthin:srcmorph-maven-plugin). |
Most of my systems work ends up as patches to other people's projects. The most complex of those is QEMU — defects in the PowerPC / pseries emulation, found while getting AIX 5.3, an operating system from 2004, to install and boot under it. Each one ships with a standalone assembly reproducer that runs bare-metal in seconds and needs no guest OS at all, so a defect can be shown to someone who has never heard of AIX. Several are not AIX-specific: a 16-bit descriptor offset in spapr_vscsi that silently corrupts guest data on transfers above 128 KiB, a standard INQUIRY vendor area returned all-zero to every SCSI guest, and slbmte / slbie disagreeing about what 32-bit mode means. The individual fixes run to a few lines each; the work was in the measuring.
Open source has interested me since I was a teenager — it started with openSUSE. Some of the other projects I have contributed to over the years:
- ⚡ Certified for working on high-voltage automotive systems (up to 1000 V AC / 1500 V DC):
Elektrofachkraft (EFffT) gem. DGUV-I 209-093 – Stufe 2E (FHV) - 🩹 Certified first aider
- 🎓 Guest lecturer on DevOps and continuous delivery in automotive software
Before moving to SRE work, I spent over a decade at IAV GmbH as a Senior Software Developer, building automotive infotainment and HMI software — the user-facing systems drivers interact with. That covered Android-based HMI development (AOSP), embedded and protocol-level work, interface design and system integration, CI/CD pipelines and test automation, performance optimization, and ASPICE process and quality standards.