Dimitris Deliakidis

Project title: Computer Vision and Machine Learning Calibration for Gentle 3D Cell Positioning in Acoustic Droplet Bioprinting

Embedded Systems Engineer Internship - UNIKLINIK RWTH Aachen University - Aachen, Germany

(2026-05 - 2026-07)

Project title: Computer Vision and Machine Learning Calibration for Gentle 3D Cell Positioning in Acoustic Droplet Bioprinting

• Designed a closed-loop calibration system to improve print reliability for a novel acoustic bioprinter: an automated pipeline learns the correlation across all printing parameters and predicts the optimal drive setting per print, applied via TwinCAT3/ADS (pyads) at sub-millisecond PLC latency, including MAX6675 thermocouple firmware (SPI, serial) for reservoir-temperature telemetry.
• Resistive heating shifts the 5.47MHz transducer's resonance, degrading a fixed drive amplitude from clean ejection and results in failure within minutes; to detect this automatically a five-stage computer-vision pipeline was made, scoring each 6×6 droplet grid into a quality score Q ∈ [0, 100] (fine-tuned YOLOv8, 96.9% F1, ρ=0.88 vs. expert grading), feeding a 4-phase online ML engine (bootstrap regression → XGBoost) plus an offline model for runtime-conditioned amplitude priors.
• Validated across 825 print cycles over 8 sessions: found print quality depends on the full multiparameter space (temperature, amplitude, runtime) rather than a fixed setpoint, raising the usable-print rate from 12% to 88% and improving droplet placement precision by 48%.

Research Assistant - University of Pennsylvania, ModLab, GRASP Laboratory - Philadelphia, PA

(2026-01)

• Developed a light displacement-based 6-axis force/torque sensor for robotic sensing, combining 2 custom PCB stacks, a compliant PDMS structure, and 24-channel optical sensing in a fingertip-scale 27 mm × 20.8 mm.
• Engineered drift-canceling calibration features and benchmarked Gradient Boosting, Random Forest, and Gaussian Process regressors, reaching sub-millimeter contact localization; sensor sensitivity improved 35% via optical redesign and ML calibration on 20/24 denoised channels at 500 Hz.

Subsystem Hardware Engineer Lead - Penn Electric Racing Formula SAE EV - Philadelphia, PA

(2024-09)

• Lead an 8-member team designing the car's Dashboard and Data Acquisition (DAQ) power systems.
• Designed and simulated the primary power management system in Altium/LTspice by integrating the team's first supercapacitor backup for safe MCU shutdown, fixing a recurring data-corruption issue from unsafe power loss; power management efficiency improved 15% (to 95%) and board area shrank 20%.
• Designed the team's first hardware-in-the-loop testing board (STM32, CAN, I2C, current sensing) to validate car boards off-vehicle, cutting validation time by two months.
• Assembled, tested, and debugged VectorNav, supercapacitor, and DAQ boards using bench equipment, catching integration issues before vehicle assembly.

Teaching Assistant - Introduction to Electrical and Systems Engineering - University of Pennsylvania - Philadelphia, PA

(2025-01)

• Led 6 hands-on labs for 80 students on MCU fundamentals, circuit components and C++ programming, reinforcing circuit design, signal processing, embedded computation, and communication protocols.
• Selected by the course professor to co-author textbook chapters that were adopted into the official course curriculum.

B.S.E. - Electrical Engineering - University of Pennsylvania (2028-05)