People · Research profile
AI systems · agent memory · trustworthy autonomy
I research how intelligent systems can remember, reason and act while preserving authority, provenance and evidence. My current work connects computer science, artificial intelligence and autonomous systems through open research and reproducible software.
Research programme
Long-lived agents need more than similarity search. They need a defensible account of where a memory came from, whether it is still valid, who may use it, what reached the model, and what happened when an interrupted agent resumed work.
My present research studies governed agent memory, execution evidence, crash continuity, structured judgment and causal attribution. Related work examines machine unlearning, generative-model control and evaluation methods for AI systems. This extends an earlier research foundation in nonlinear control, autonomous ground vehicles, soft robotics, system modelling and field experimentation.
Current research
Open papers, datasets, benchmark results and implementations are published so that claims can be inspected independently.
Technical paper · dataset · interactive results · 2026
Two-axis evaluation of memory authority and operational continuity, covering 700 seeded integrity trials, installed-product crash checks and controlled four-arm replay.
Research article · 2026
Architecture and claim boundaries for memory authority, lifecycle, delivery evidence and reconstructable agent runs.
Read on Hugging Face ↗Retrieval study · 2026
An exploratory AtMem–Jev study separating candidate retrieval, ranking judgment and final memory authorization.
Read the study ↗Evaluation infrastructure
An open benchmark surface for comparing agent-memory systems across safety, governance and retrieval behavior.
Open the benchmark ↗Open-source research system
AtMem is an open-source memory control plane and Agent Black Box for AI agents.
The project treats relevance and authorization as separate questions. It combines canonical memory, governed retrieval, context-delivery evidence, lifecycle controls and execution traces so researchers can examine what an agent remembered and how that memory influenced a run.
Selected academic work
Selected book, journal and conference publications. The complete and current record is maintained on Google Scholar.
Springer International Publishing · book
Autonomous Robots 42(3), 553–568
IFAC-PapersOnLine 49(16), 469–474
IEEE/ASME International Conference on Advanced Intelligent Mechatronics
7th German Conference on Robotics
Doctoral research
PhD in Mechatronics. Research on path tracking for agricultural and off-road vehicles, robust control under wheel slip, modelling, simulation and field validation.
Research formation
MSc in Information and Automation Engineering. Research included adaptive position control for pneumatic soft robotic actuators and dynamic and kinematic modelling of robotic systems.
Research contact
For questions about AtMem, benchmark methodology, datasets or related AI-agent research.
Google Analytics measures visits on public pages. Analytics cookies are on by default. Turn off analytics cookies to continue with cookieless measurement.