Portfolio
This page presents a portfolio of my engineering work at the intersection of robotics, AI, and embedded systems. Each summary outlines the problem tackled, the approach taken, and the key technologies applied. I focus on practical innovation, rapid prototyping, rigorous testing, and secure deployment. Read on to see how I push the limits of small, resource‑constrained platforms to deliver reliable, high‑impact solutions.
NLP Powered Drone Control
US Navy Research Project
Leveraged BERT‑based language models to translate spoken or typed instructions directly into precise flight commands, giving operators an intuitive “chat‑to‑fly” interface and richer situational awareness. The effort covered the end‑to‑end build of custom UAV platforms, a hardened ground‑to‑air communications stack, and deployment of the NLP pipeline on resource‑constrained embedded hardware. Security assessments throughout the stack uncovered—and resolved—critical hardware and software vulnerabilities, ensuring robust field operation.
Technologies & skills
- Threat modeling & penetration testing
- Python │ PyTorch • Transformers (BERT fine‑tuning, INT8 quantization)
- Secure telemetry: LoRa, WPA2/3 Wi‑Fi, VLAN segmentation
- Embedded Linux on Raspberry Pi 4/5, Jetson Orin, RP2040
- Docker • k3s edge orchestration
- 3‑D printing & carbon‑fiber fabrication
- ArduPilot • MAVLink • BLHeli32 ESCs
Turbine Powered Drone
Collaboration with AERo Department
Co‑developed with Cal Poly Pomona’s Aerospace Engineering Department, this platform trades electric rotors for twin JetCat P60 micro‑turbine engines; delivering > 13 kg combined thrust and true fixed‑wing‑class cruise speeds while retaining multirotor agility for VTOL. The result is a long‑range, high‑payload drone suited for rapid aerial mapping, heavy‑lift logistics, and propulsion research.
Technologies & skills
- JetCat P60‑SE turbines • kerosene start sequence
- Hybrid avionics: ArduPilot VTOL flight stack + real‑time fuel‑flow PID
- High‑temp composites (carbon‑fiber/Kevlar + ceramic thermal barriers)
- Aerodynamics & propulsion: XFOIL/AVL sizing, Ansys Fluent CFD, thrust‑stand validation
- On‑board data logger (Jetson Orin Nano)
- Ground‑station dashboards (Flask + Socket.IO) for live EGT, RPM, fuel burn
- Collaborative design reviews with Aero Dept.
Reconfigurable CubeSat Clusters
CPP Engineering Showcase 2025 – First Place
Designed a modular compute‑and‑communications stack that turns a swarm of CubeSats into one elastic “orbital data‑center.” Each 1 U satellite carries a custom board pairing a low‑power CPU with a Xilinx FPGA that supports on‑orbit partial reconfiguration. A lightweight k3s/Kubernetes control plane lets workloads migrate between nodes in seconds, creating self‑healing compute clusters and redundant down‑links; at roughly 1 % of the cost of today’s space‑qualified servers.
Technologies & skills
- Cost‑down BOM analysis & design‑for‑manufacture (DFM)
- Xilinx Artix‑7 / Zynq MPSoC · Vivado Partial Reconfiguration
- k3s (Kubernetes on edge devices) · Docker · Helm CI/CD
- Fault‑tolerant mesh networking, QoS, forward‑error correction
Submersible Drone
NASA MINDS 2025 – Exceptional Experimental Design
Engineered a dual‑medium UAV/ROV that transitions seamlessly between aerial flight and underwater propulsion. AQUAD combines lightweight EDF thrusters, a pressure‑tolerant frame, and adaptive control firmware to enable aerial insertion, subsurface maneuvering, and vertical extraction on a single charge. Developed for the NASA MINDS Competition, it targets ocean‑world exploration (e.g., Europa) while doubling as a rugged platform for coastal and flood‑zone search‑and‑rescue.
Technologies & skills
- Rapid prototyping, interdisciplinary team leadership
- EDF propulsion (50 mm)
- Python / C++ · MAVLink · real‑time dashboards (Flask + Socket.IO)
Tesla Coil Drone
NASA MINDS 2024 – Honorable Mention
Developed a dual‑purpose system that uses paired Tesla coils, tuned in resonance, to beam power and low‑frequency data over short‑to‑medium ranges without physical conductors. The work demanded deep electromagnetic‑compatibility (EMC) engineering to protect sensitive avionics and sensor payloads operating in kilovolt‑level near‑fields, as well as bespoke shielding and filtering strategies for components unsuited to high‑EMF environments.
Technologies & skills
- High‑frequency, high‑voltage power electronics
- EMI/EMC mitigation: Faraday cages, mu‑metal shielding, opto‑isolated I/O
- Field testing with oscilloscope/VNA
Aerial Drone Docking
NASA MINDS 2024 – Finalist
Built a heavy‑lift “mothership” UAV that serves as an airborne charging hub and data relay for multiple micro‑drones. Precision vision‑based docking lets each satellite drone land, top‑off its batteries, off‑load sensor data, and even share GPU cycles; extending mission endurance from minutes to hours and turning the swarm into a distributed compute cluster in the sky.
Technologies & skills
- Autonomous docking: OpenCV, visual‑servo PID loops
- ArduPilot‑MAVLink swarm control
- 5 GHz Wi‑Fi mesh + LoRa telemetry
- Distributed compute: Jetson Xavier NX / Orin Nano, k3s Kubernetes, Docker
- Custom carbon‑fiber frame
- Real‑time dashboards (Flask + Socket.IO)
Transformer Drone
NASA MINDS 2024 – Semi-Finalist
Engineered a morphing multirotor whose thruster pods pivot and slide on servo‑driven rails, letting the airframe shift from a wide‑span “cruise” mode to an ultra‑compact “crawl” footprint in seconds. The variable geometry optimizes lift‑to‑drag for long‑range flight yet tucks the rotors close to the fuselage to slip through doorways, culverts, and industrial pipe racks. Modular motor arms and swappable prop modules also make the architecture easy to down‑scale for micro‑UAV applications.
Technologies & skills
- Agile hardware iteration, design‑for‑manufacture cost analysis
- Adaptive flight control: ArduPilot · custom PID re‑tuning on‑the‑fly
- BLHeli32 ESCs, 4‑S/6‑S Li‑ion packs
- Rapid prototyping: CNC‑milled CF panels & Resin Components
