Skip to content
View Dimitrios-Kafetzis's full-sized avatar

Block or report Dimitrios-Kafetzis

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
Dimitrios-Kafetzis/README.md

Hi, I'm Dimitrios Kafetzis 👋

Dimitrios Kafetzis — Electronics & Software Engineer, Systems Architect, Researcher

Website Blog SynapticOS Google Scholar ResearchGate arXiv LinkedIn Email


  • 🎓 PhD candidate in AI & Networks at the Athens University of Economics and Business, Dept. of Informatics, advised by Prof. Iordanis Koutsopoulos. I work on partitioning DNNs and LLMs for low‑latency inference across resource‑constrained edge devices, and on SDN/SDR for tactical mobile ad hoc networks.

  • 🏭 Embedded systems architect in the maritime industry: custom Yocto Linux distributions, wireless sensor networks on Zephyr and FreeRTOS, and systems built to IEC 60945, IEC 61162 and IEC 61162-450, IEC 61174, IEC 62923, IEC 62443 and IACS E26/E27.

  • 🧠 Building SynapticOS, an AI‑native RTOS for Cortex‑M33 + NPU microcontrollers built on Zephyr. Inference is the primary workload, not a library bolted onto a scheduler. Three papers on its architecture are on arXiv.

  • ⚙️ Upstream contributor to Zephyr RTOS: fixes across the CMake/sysbuild build system, NXP MCX board and devicetree support, flash and USB drivers, cbprintf and driver tests. See my merged and open pull requests. Recognised by The Linux Foundation with the Zephyr Technical Contributor badge.

  • 🧪 Distributed Inference Testbed, the experimental platform of my PhD: five Raspberry Pi 4 boards on an isolated 802.11ac network, an edge server, and the NNPM web application that runs every experiment end to end with controlled network and compute conditions, per-layer profiling, power traces and partition optimisation. Runs without hardware in local mode.

  • 📄 Latest research: HeadSplit, attention‑head partitioning and inter‑layer speculation for distributed LLM inference at the edge (IEEE SPAWC 2026), and LLM partitioning for low‑latency edge inference (IEEE/IFIP WiOpt 2025).

  • 🔧 15+ years across the hardware/software stack: from PCB design and FPGAs to ML and edge–cloud architectures. I have led R&D teams in the maritime, defence and gaming industries.

  • 📫 Open to research collaborations and interesting embedded/edge‑AI problems: dimitrioskafetzis@gmail.com · LinkedIn


🚀 What I'm building right now

SynapticOS HeadSplit Fieldproof EdgeNN



🖥️ Projects

Edge AI & Embedded Systems
ProjectStarsWhat it is · Tech
SynapticOS Stars AI‑native RTOS for Cortex‑M33 + eIQ Neutron NPU (NXP FRDM‑MCXN947), built on Zephyr
EdgeNN Stars Zero‑allocation C11 inference library for Cortex‑M/A: Dense, Conv, LSTM, GRU, Transformer with INT8
EdgeOrchestrator Stars Distributed, adaptive resource orchestrator for AI workloads on embedded Linux clusters
edge‑inference‑rt Stars Memory‑aware runtime for partitioned Transformer inference on constrained devices
axon Stars Header‑only sensor‑fusion framework: EKF, UKF, complementary filters, Mahalanobis outlier gating
Distributed DNN / LLM Inference (research code)
ProjectStarsWhat it is · Tech
Distributed Inference Testbed Stars PhD testbed: partitioned DNN and LLM inference across Raspberry Pi 4 devices and an edge server, driven by the NNPM web app; tc‑netem shaping, load generation, INA260 power, RR/SJF delay model and partition optimisation
HeadSplit Stars Attention‑head partitioning + inter‑layer speculation for distributed LLM inference at the edge
Distributed Transformer Inference Simulator Stars Discrete‑event simulator for resource‑aware decoder‑only Transformer inference across edge clusters
DNN Partitioning & Offloading Testbed Stars Raspberry Pi testbed for layer‑level DNN partitioning: Greedy, SA, GA; power and latency profiling
Video QoE Forecasting Stars Linear, LSTM, GRU and Transformer forecasters for video QoE from network metrics
Networks & Security
ProjectStarsWhat it is · Tech
Ryu SDN + Snort Reactive Firewall Stars OpenFlow controller app that blacklists and drops traffic on Snort IDS alerts
SDN‑MANET Firewall Simulators Stars Simulators for reactive firewalling in software‑defined mobile ad hoc networks
RNN vs Transformer DDoS Detection Stars Systematic comparison of RNN and Transformer detectors for transport‑layer DDoS on IoT/5G gateways
Robotics, Maritime & Desktop Tools
ProjectStarsWhat it is · Tech
Fieldproof Stars Multi‑domain robotics emulation (UGV, UAV, USV): unmodified ROS 2 stacks against Gazebo, scenario‑driven tests
NMEA Simulator X Stars Cross‑platform NMEA 0183 and Signal K vessel data simulator over serial, TCP, UDP and WebSocket
Deucalion Stars Back up an Ubuntu machine to an external drive and rebuild it on new hardware or a new release
Nature for Cooling Stars Screening‑level decision support for nature‑based urban cooling solutions, developed for Criterra
Upstream open‑source contributions
ProjectPRsWhat I contributed · Areas
Zephyr RTOS Merged PRs
Open PRs
Zephyr Technical Contributor badge
Build‑system and sysbuild fixes, NXP MCXN/MCXA board and devicetree support, Renesas RA flash driver, USB CDC ACM logging, cbprintf bounds, UART async test stability, OpenTitan PLIC
Where it started (2013)
ProjectStarsWhat it is · Tech
Snake on Stellaris Stars Snake written in C for the TI Cortex‑M3 Stellaris evaluation board
Slot Deck for iPhone Stars The classic slot deck game in Objective‑C on Cocoa

📚 Publications

Citations h-index i10-index Google Scholar, October 2026

Year Title Venue Links
2026 HeadSplit: Attention Head Partitioning and Inter‑Layer Speculation for Distributed Inference at the Edge
D. Kafetzis, I. Koutsopoulos
IEEE SPAWC 2026 Code
2026 SynapticOS: An Inference‑First Runtime Architecture for Neural Processing Units on Resource‑Constrained Microcontrollers
D. Kafetzis
arXiv preprint arXiv · Code
2026 Inference Pipelines as Operating‑System Objects: Priority Scheduling and Constant‑Footprint Streaming for Microcontroller Neural Inference
D. Kafetzis
arXiv preprint arXiv · Code
2026 Cross‑Core Inference Offload as an Operating‑System Service on Dual‑Core Microcontrollers
D. Kafetzis
arXiv preprint arXiv · Code
2025 Video Quality Monitoring for Remote Autonomous Vehicle Control
D. Kafetzis, N. Fotiou, S. Argyropoulos, N. Nasreddine, I. Koutsopoulos
IEEE MeditCom 2025 PDF
2025 Large Language Model Partitioning for Low‑Latency Inference at the Edge
D. Kafetzis, R. Khalili, I. Koutsopoulos
IEEE/IFIP WiOpt 2025 arXiv · PDF · Code
2024 An Experimental Platform for AI Model Partitioning on Resource‑Constrained Devices (Demo)
D. Kafetzis, I. Koutsopoulos
ACM MobiHoc 2024 DOI · Code
2024 DNN Partitioning and Inference Task Offloading in 6G Resource‑Constrained Networks
D. Kafetzis, I. Koutsopoulos
EuCNC & 6G Summit 2024 DOI · PDF
2023 Joint Controller Placement and Intrusion Detection System Enablement in Software Defined Mobile Ad Hoc Networks
D. Kafetzis, I. Koutsopoulos
European Wireless 2023 PDF
2023 Revisiting the OLSRv2 Protocol Optimization in SDN‑enabled Tactical MANETs
I. Fourfouris, M. Karaliopoulos, D. Kafetzis, G. Vardoulias, L. Georgiadis, S. Vassilaras, I. Koutsopoulos
IEEE MILCOM 2023 DOI
2022 Software‑Defined Networking Meets Software‑Defined Radio in Mobile ad hoc Networks: State of the Art and Future Directions
D. Kafetzis, S. Vassilaras, G. Vardoulias, I. Koutsopoulos
IEEE Access DOI · PDF
2020 UAV‑assisted Aerial Survey of Railways using Deep Learning
D. Kafetzis, I. Fourfouris, S. Argyropoulos, I. Koutsopoulos
IEEE ICUAS 2020 DOI · PDF
🔬 Funded research projects I have worked on
Project Programme Role
TARGET‑X (2024–2025) Horizon Europe SNS‑JU Work‑package lead: ML‑driven QoE prediction for remote automotive tele‑operation
SMOTANET (2020–2024) EDIDP, with Intracom Defence Work‑package lead: cybersecurity and network forensics in software‑defined tactical MANETs
LEMONADE (2019–2020) HFRI Delay‑tolerant mobile networks and DNN partitioning for edge AI
EREVOS (2019–2020) H2020 / ESMERA Computer‑vision UAV navigation for railway inspection on NVIDIA Jetson Nano
DAIAD (2014–2016) EU FP7 Embedded systems and BLE hardware for smart water monitoring
LAROS (2015–2016) Prisma Electronics Maritime vessel monitoring platform
WPSS (2012) CERN ATLAS Wireless personnel safety and fall‑detection system
UARP v2 (2012–2013) University of Leeds FPGA‑based ultrasound research platform
🏆 Awards
  • Innovation Radar Prize, European Commission (2017), for the Amphiro b1 product developed in DAIAD
  • Technical Achievement Award, Lloyd's List Greek Shipping Awards (2015), for LAROS
  • Postgraduate Studies Scholarship, University of Leeds (2012)
  • CPC Raspberry Pi Hackathon winner (2012)
  • Best Embedded Systems Project, Farnell competition (2011)
  • Best Electronics Design Project, Circuit Analysis and Design competition (2010)

🧰 Toolbox

Languages

Embedded & Hardware

AI / ML

Networks, Protocols & Security

Robotics, Tooling & Cloud


📈 GitHub stats

GitHub stats Top languages

Contribution streak


🎓 BEng Electronics & Nanotechnology (First Class Honours) and MSc Embedded Systems Engineering, University of Leeds · PhD candidate, AUEB
📍 Athens, Greece & Hamburg, Germany · Teaching Fellow at AUEB (Computer Networks, C++, Databases) · Thesis supervisor for BSc and MSc students

Pinned Loading

  1. SynapticOS SynapticOS Public

    AI-native RTOS for Cortex-M33 + eIQ Neutron NPU, built on Zephyr. Target: NXP FRDM-MCXN947.

    C 5

  2. zephyrproject-rtos/zephyr zephyrproject-rtos/zephyr Public

    Primary Git Repository for the Zephyr Project. Zephyr is a new generation, scalable, optimized, secure RTOS for multiple hardware architectures.

    C 16.9k 10.3k

  3. HeadSplit HeadSplit Public

    Reference implementation of HeadSplit (IEEE SPAWC 2026): attention head partitioning and inter-layer speculation for distributed LLM inference on edge devices.

    Python

  4. EdgeNN EdgeNN Public

    Lightweight, zero-allocation C11 library for neural network inference on ARM Cortex-M/A microcontrollers. Supports Dense, Conv2D, LSTM, GRU, Multi-Head Attention, and 18 operators total with INT8 q…

    C

  5. NMEA_Simulator_X NMEA_Simulator_X Public

    Free, open-source NMEA 0183 and Signal K data stream simulator for Windows, macOS and Linux. C++20 and Qt 6.

    C++ 1

  6. RNN-Transformer-DDoS-Detection-Comparison RNN-Transformer-DDoS-Detection-Comparison Public

    A systematic comparison of RNN and Transformer architectures for detecting transport-layer DDoS attacks in resource-constrained IoT/5G environments.

    Python 1