Šimon Hudínek

CVS emission measurements at Škoda Auto are humidity-sensitive, so I designed the measurement campaign, instrumented the test cell, and built a validated Python thermodynamic model with an operator GUI for pre-checking run conditions. Operators now flag condensation-prone runs before they happen, cutting invalid tests and rework. The panel here is that pre-check tool rebuilt in the browser: set ambient conditions, dilution factor, and fuel, and a Magnus-formula psychrometric model computes the dilute-mixture dewpoint against the tunnel wall temperature.

Diesel-generator aftertreatment needs precise AdBlue dosing: too little and NOx escapes, too much and ammonia slips through. For my master's thesis I coupled a 1D SCR catalyst model with an NMPC dosing strategy and validated it against the XMR monolith reactor simulator — no hardware safety net. The controller holds NOx conversion across a full stationary-point load cycle while keeping NH3 slip under 1 ppm, where the baseline PID peaks at 262 ppm. The panel here replays that benchmark — outlet NOx and NH3 slip for both controllers, scrubbable and playable over the 420-minute cycle.

A physical multi-tank water plant needed a modern control system, so I built the full stack: a first-principles plant model, a nonlinear MPC with EKF state estimation, NI-DAQmx data acquisition, and a PySide6 operator application. The result is stable closed-loop control on the real rig with constraint handling. The panel here is that operator app — replaying a precomputed closed-loop run at 6× real time.

Situation

Everyday problems worth engineering: healthy lighting rhythms and hydration tracking.

Task

Design small, self-contained embedded systems that run reliably without cloud dependencies.

Action

Built an ESP32-based circadian lighting cabinet controlled by a local server, and a sensor-equipped ESP32 smart bottle.

Result

Daily-driver devices demonstrating full-stack embedded skills: firmware, electronics, enclosure, and local networking.