Martin Le

Research software engineer — wireless systems

Teaching Wi-Fi networks to pick the right speed, one packet at a time.

I design and deploy machine-learning algorithms — mostly multi-armed bandits — that let real 802.11 hardware decide how fast to transmit, and I build the testbeds that prove the decisions hold up off the whiteboard.

exploration convergence on best arm →
01 — Summary

Six years, one testbed, a lot of packet captures.

from the lab notebook

I'm an Information Systems Engineering graduate and PhD candidate specializing in wireless systems, workflow automation, and machine-learning-driven optimization for Wi-Fi networks. Most of my working life has been spent moving algorithms out of notebooks and onto real hardware — building the automated experimentation platforms, Python tooling, and monitoring frameworks that make that move survivable.

I care about the unglamorous middle of research: reproducible pipelines, testbeds that don't need babysitting, and results that hold up when you swap the whiteboard for a shielded chamber — and then for the open air.

Role
Research Software Engineer
Institution
Technische Universität Braunschweig
Focus
Multi-armed bandits, Wi-Fi rate adaptation, testbed automation
Languages spoken
German, English, Korean, Vietnamese
02 — Focus

What I actually spend my days on.

Algorithms

Bandits that pick a rate

Thompson Sampling and its variants — RAGTS, CTS, UTS, CUTS — reworked as rate-adaptation logic for IEEE 802.11ac, benchmarked head-to-head against Minstrel-HT rather than against each other in isolation.

Infrastructure

Testbeds that don't lie

Automation and monitoring frameworks spanning shielded RF chambers and over-the-air deployments — the difference between a result that looks good in simulation and one that survives real interference.

Delivery

From prototype to hardware

Python software for configuring, monitoring, and running Wi-Fi systems end to end — taking an ML algorithm from research prototype to something that runs unattended on commercial routers.

03 — Experience

The trace, traced out.

2019 — 2025

PhD Student / Research Associate

Technische Universität Braunschweig
  • Designed and built Python software for configuration, monitoring, and automation of Wi-Fi communication systems
  • Took machine-learning algorithms from research prototype to deployment on experimental hardware
  • Built reproducible software pipelines for large-scale experimentation and data analysis
  • Established and maintained the group's experimental testbed for performance evaluation
  • Supervised software engineering projects and mentored Master's students
SupraCoNeX (BMFTR) · +18% throughput
2019

Student Assistant

Technische Universität Braunschweig
  • Set up and optimized throughput for Wi-Fi networks using commercial off-the-shelf routers from various manufacturers
2018 — 2019

Student Assistant

Technische Universität Dresden
  • Wrote Python and Bash scripts to control and monitor Wi-Fi router parameters
  • Ran automated experiments and measurements on live Wi-Fi networks
2016 — 2018

Student Assistant

Fraunhofer Institute for Machine Tools and Forming Technology
  • Designed and built a MATLAB-based GUI for monitoring and controlling simulations
2015 — 2016

Internship

Fraunhofer Institute for Machine Tools and Forming Technology
  • Automated document generation in Word and Excel using VBA
  • Built a Python GUI for monitoring and controlling simulations
  • Wrote scripts for experimental data evaluation and visualization in Python and MATLAB
Education — Diploma, Information Systems Engineering, Technische Universität Dresden (2010–2019). Thesis: "Wi-Fi Interference Management: Periodic Remote Selection of Least Interfered Bands," a channel-selection algorithm built with linear programming and graph theory.
04 — Selected work

Projects worth a second look.

2025 · IEEE Transactions on Communications

WiFi-CUTS

Rate adaptation with cascaded unimodal multi-armed bandits, validated in an IEEE 802.11ac testbed rather than simulation alone — the core algorithmic contribution of my PhD.

Journal paper · early access
2025 · ISIT

Wireless Wanderlust

Semi-supervised indoor positioning from Channel State Information, using an autoencoder paired with a fully connected position predictor.

Junior prize · team of three
2012 · TU Dresden

Event Management System

A web-based party event management platform — user accounts, database, and calendar integration — carried through design, build, test, and client acceptance.

Full-lifecycle software project
2010 · RoboLab

Line-Following Robot

A Lego Mindstorms robot that reads solid, dashed, curved, and zigzag lines while avoiding obstacles — my first taste of control loops and edge cases.

1st place · team of four
2024 · MobiCom

Open Resource Control API

Contributed to an open-source resource control API for real IEEE 802.11 networks, presented at ACM MobiCom with Sankalp P. Pawar and collaborators.

Conference paper · co-author
05 — Publications

What's been peer-reviewed.

  • 01

    M. Le, B. Peng, and E. A. Jorswieck, "WiFi-CUTS: Rate Adaptation with Cascaded Unimodal Multi-Armed Bandits in IEEE 802.11ac Testbed Experiments,"

    IEEE Transactions on Communications (Early Access), 2025 · doi: 10.1109/TCOMM.2025.3628731

  • 02

    M. Le et al., "Multi-Armed Bandits in IEEE 802.11ac: Efficient Algorithms and Testbed Experiments,"

    IEEE Intl. Workshop Technical Committee on Communications Quality & Reliability (CQR), Seattle, WA, 2024, pp. 7–12 · doi: 10.1109/CQR62340.2024.10705888

  • 03

    S. P. Pawar et al., "Open-source Resource Control API for real IEEE 802.11 Networks,"

    Proc. 30th Annual Intl. Conf. on Mobile Computing and Networking (ACM MobiCom), New York, 2024, pp. 1866–1873 · doi: 10.1145/3636534.3697314

06 — Toolkit

What's on the bench.

Languages
  • Python
  • Bash
  • Java
  • MATLAB
Data & ML
  • PyTorch
  • scikit-learn
  • Polars / Pandas
  • NumPy
Tooling
  • Git
  • Jupyter / Marimo
  • Matplotlib / Seaborn
  • REST APIs
Systems
  • Linux
  • OpenWrt
  • macOS
  • Windows
German — native English — fluent Korean — upper intermediate Vietnamese — upper intermediate