I'm a Master's student in Aerospace Engineering at Technische Universität Darmstadt, Germany — working at the intersection of aviation and software.

My focus: Machine Learning for aerospace systems, Cockpit UI/UX, and Human-Machine Interaction (HMI). I build physics-grounded tools and pilot-centric interfaces — software that respects the engineering underneath it.

🎯 Currently open to Master's Thesis opportunities, student jobs, and research collaborations in Aerospace Engineering, ML, and Data Science.

Mar – Apr 2026  ·  Personal Project  ·  GitHub ↗  ·  Details ↗

Full-stack Python desktop application for space mission planning. Precision astrodynamics engine (Hohmann, Bi-Elliptic, Phasing transfers computed analytically) paired with a neural network surrogate model — wrapped in a PySide6 mission control dashboard with live animated 2D orbital trajectories.

The ML insight: all delta-v formulas reduce to circular orbital velocities v_c = √(μ/r) — so one model covers Earth, Moon, and Mars with no body flags, no lookup tables. Trained on 200,000 synthetic scenarios generated by the physics engine itself. Input range capped at 400,000 km to keep predictions inside the model's training distribution.

Oct – Dec 2025  ·  Personal Project  ·  Details ↗

End-to-end ML pipeline predicting engine Remaining Useful Life (RUL) on the NASA C-MAPSS dataset. The headline result isn't accuracy — it's safety: a custom Asymmetric Loss Function penalises overestimation 10× harder, cutting dangerous safety violations from 12.41% → 1.09%.

LSTM (64 → 32 units, Dropout 0.2) with a 30-cycle sliding window captures degradation velocity, not just instantaneous sensor values. K-Means cluster similarity features encode operating mode awareness and dominate feature importance at ~37%. Architecture tuned with Keras Tuner.

Apr – Sep 2025  ·  FSR — TU Darmstadt  ·  Details ↗

Modular cockpit UI for single-pilot operations — PFD, MFD, ECAM systems pages, flight controls panel, ATC management, and backup controls — running live in the FSR simulator dome at TU Darmstadt. Multi-threaded UDP layer streams real-time X-Plane 12 telemetry at 50 Hz with < 20 ms end-to-end latency. Custom AnimatedSlider components with discrete detents (flaps, speed brakes, elevator trim) reused across all slider-based controls. Awarded grade 1.3.

Oct 2024 – Mar 2025  ·  Boeing × FSR — TU Darmstadt  ·  Details ↗

30 years of major runway incursion events analysed under the SURFIA framework, structured against RTCA DO-323 (94% of operational requirement categories covered). Permutation-importance scoring ranked contributing factors by marginal effect on collision probability. Findings presented directly to Boeing safety engineers at Boeing facilities. Awarded grade 1.7.

Apr – Sep 2024  ·  TU Darmstadt  ·  Details ↗

Mechatronic mowing system with active stabilisation control. PID controller designed and tuned in MATLAB/Simulink, lead-screw mechanism CAD-modelled in Fusion 360 and 3D-printed on a Prusa i3 MK3. System integration managed via agile SCRUM. All stability specs met on first hardware test.

📍 Darmstadt, Germany  ·  🗣️ English (Fluent)  ·  German (B1/B2)