M.S. Computer Science @ Embry-Riddle · Open to SWE roles

CARMEN LUCCA
DIMARIO

I work at both ends of the stack. At the low level, that means compiler toolchains and ROS2 autonomy running on embedded hardware. Higher up, it means iOS apps, Go blockchains, and web platforms.

See the workDownload résuméGitHub
Selected Work

Things I've built

A post-quantum blockchain, an AI checkride examiner, and the middleware flying a drone. Each one was shipped, tested, or published.

Systems / Cryptography/2026

Dilithium

Post-quantum cryptocurrency, written from scratch in Go

A peer-to-peer proof-of-work blockchain that is quantum-safe from block zero. Every transaction is signed with CRYSTALS-Dilithium Mode3, the NIST FIPS 204 standard, so there is no migration to make when quantum computers arrive. Node, wallet CLI, and CPU/GPU miner all built from first principles.

  • NIST Level 3 (192-bit) quantum-safe signatures via CRYSTALS-Dilithium Mode3
  • Competitive PoW mining across a live P2P network with automatic block and transaction sync
  • REST API, UPnP port forwarding, and cross-platform builds for Linux, macOS, and Windows
GoCRYSTALS-DilithiumCloudflare CIRCLP2PSHA-256
dilithiumcoin.comSource
iOS / Applied AI/2025 — present

PFactor

AI oral-exam simulator for FAA checkride prep

Student pilots practice their checkride oral exam by actually talking to an adaptive AI examiner. It asks follow-ups on weak answers, scores each response the way a real DPE would, and delivers per-category debriefs. No typing, no multiple choice.

  • Adaptive examiner engine scores responses 1–5, generates targeted follow-ups, and mirrors real DPE behavior
  • ACS-aligned question banks covering 6 FAA certificates
  • Serverless Convex backend handling Sign in with Apple, StoreKit 2 subscriptions, and graceful offline degradation to local scoring
SwiftSwiftUIConvexStoreKit 2LLM APIsNext.js
pfactor.app
Robotics / Embedded/2024 — 2026

AIRHOUND

Vision-based UAV target tracking, middleware lead

Built the middleware that turns a camera feed into flight commands. A modular four-node ROS2 pipeline converts RF-DETR detections into yaw commands via pinhole camera intrinsics and publishes them to a PX4 flight controller over Micro XRCE-DDS. Work published at SPIE Defense + Security 2026.

  • Diagnosed DDS image-transport serialization as the pipeline bottleneck and replaced it with direct V4L2 capture, cutting mean detection latency from 234 ms to 58.8 ms and eliminating dropouts at 1280x720
  • Optimized RF-DETR with NVIDIA TensorRT FP16 for real-time inference at 15 FPS on a Jetson Orin
  • Designed the detection-to-actuation contract so perception and control teams could iterate independently
C++PythonROS2RF-DETRTensorRTPX4Jetson OrinMicro XRCE-DDS
PaperFull systemMy middleware

Also built

Robotics / Embedded/2025 — 2026

RoboRacer

Software lead, F1Tenth competition autonomy stack

A 1/10-scale autonomous race car, built from bare components into a competition vehicle. I led the software: mounting the compute, sensors, and motor controller, bringing up the full ROS2 stack on the onboard computer, and taking the autonomy pipeline from sensor input through planning to actuation under real race conditions.

  • Brought up the onboard compute and sensor stack from scratch: LiDAR, IMU, and motor controller integrated over a single ROS2 graph on embedded Linux
  • Tuned the perception-to-actuation loop for the latency budget a racing line demands, where a late control command means a wall instead of an apex
  • Debugged across the hardware-software boundary, the same failure surface as AIRHOUND: when the car misbehaves, the cause is as often a sensor timing or wiring fault as it is the planner
  • Validated the end-to-end stack under real competition conditions, where recovery has to be automatic and there is no operator in the loop
ROS2PythonC++SLAMLiDAREmbedded Linux
Research / Reinforcement Learning/2026 — present

SEMBAS-RL

Graduate research, autonomous driving policy assurance

Reinforcement learning policies fail in ways their training curves never show. SEMBAS-RL extends an adaptive boundary scanner to RL policy evaluation, mapping the 3D scenario space to find where a learned driving policy stops being reliable, then turning that geometry back into concrete retraining proposals.

  • Trained and evaluated thirteen driving policies across two algorithms (PPO, SAC) and five curriculum strategies over 5M+ training steps
  • Extended SEMBAS, a three-phase adaptive boundary scanner (Latin hypercube seeding, binary bisection, boundary walking), to RL policy evaluation, wired to the simulator as a socket service returning deterministic PASS/FAIL labels
  • Built three analysis agents (policy down-selection, failure-mechanism classification, boundary-aware retraining planning) that convert boundary geometry into ranked retraining proposals, closing the loop from evaluation back into training
PythonPyTorchStable-Baselines3GymnasiumPPO / SACCARLoS
Source
Machine Learning/2026

SiT Trajectory Prediction

Pedestrian forecasting from LiDAR on an HPC cluster

Pedestrian trajectory prediction on the SiT dataset, fusing LiDAR point-cloud features with kinematic and ego-motion signals. Trains Decision Tree, XGBoost, and a Social Transformer, benchmarked by ADE/FDE.

  • 45-feature pipeline: 34 kinematic, 4 ego-motion, and 7 LiDAR-derived features
  • Reproducible PBS/Torque job scripts targeting ERAU's Vega cluster, with a Colab GPU path
PythonPyTorchXGBoostLiDAR / PCDPBS/Torque
Source
Systems / TUI/2026

amuse

Terminal player for Amazon Music

Amazon deprecated its desktop app and DRM'd the streams, so amuse wraps the browser tab instead: OS-native media session control (Windows SMTC, Linux MPRIS), a live FFT spectrum visualizer driven by system audio loopback, and album art rendered as true-color ANSI half-blocks.

  • WASAPI and PulseAudio/PipeWire loopback capture feeding a real-time FFT visualizer
  • Per-process volume control through the Windows mixer via pycaw; native MPRIS volume on Linux
  • No scraping and no unofficial APIs; it drives the same interfaces your media keys do
PythonWASAPIMPRIS / D-BusFFTTextual
Source
BMG Aviation screenshot
Freelance/2025

BMG Aviation

Aviation services client site

Marketing site designed and shipped end to end for an aviation services company, built for speed and search visibility.

Next.jsTypeScriptTailwind
Live site
Pocopson Vet Station screenshot
Freelance/2026

Pocopson Vet Station

Veterinary practice client site

Full redesign and build for a veterinary practice, from information architecture through deployment.

Next.jsTypeScriptTailwind
Live site
Experience

Where I've worked

Embedded Software Development Intern

Garmin

May 2026 — August 2026

  • Implemented Clang compiler support as a replacement for legacy ARMcc toolchains
  • Extended embedded library support to 64-bit architectures, resolving data type alignment and pointer-width issues to broaden platform compatibility
  • Developed a system test suite for a proprietary avionics communication interface, validating real-time engine metric transmission and protocol correctness under embedded constraints
CC++ClangARMccGNU ASMAvionics protocols

RISE Intern / CO-OP Participant

JRC Integrated Systems

June 2025 — May 2026

  • Upgraded the PETSc linear algebra library across a major version boundary to enable GPU acceleration, reducing complex computation time
  • Analyzed an advanced C/C++ codebase to identify differences in Linear Energy Transfer (LET) calculations through various substrates
  • Rebuilt the Accuro test suite from hard-coded shell scripts into a Python framework with parallel execution and selectable test subsets, cutting a multi-hour run to roughly 25 minutes
CC++PETScNumerical methodsTest infrastructure

Full Stack Development Intern

ivDash

May 2024 — August 2024

  • Led development of an admin dashboard consolidating functionality for doctors, nurses, and product providers across a digitalized healthcare distribution platform
  • Built interactive web interfaces in Next.js against an existing Symfony PHP backend
  • Streamlined Auth0 JWT authentication to work across both the Symfony PHP and Next.js applications
Next.jsTypeScriptSymfonyPHPAuth0 / JWT
About

Background

Finishing a B.S. and M.S. in Computer Science at Embry-Riddle while running Division II cross-country. Two peer-reviewed publications, an FAA Part 107 certificate, and a habit of building the thing rather than reading about it.

Languages

CC++SwiftGNU ASMTypeScriptJavaScriptGoJavaPythonSQLBash

Embedded & Robotics

ROS2PX4STM32Jetson OrinClangARMccTensorRTUNIX

Web & Mobile

ReactNext.jsSwiftUIConvexSymfonyLaravelTailwind

Practice

AgileScrumTest infrastructureTechnical writing

Publications

Malarchick, R., DiMario, C., et al. "Predictive Target Pursuit for Autonomous UAVs Using RF-DETR with Depth-Aware State Estimation and Physics-Informed Trajectory Prediction."

Proc. SPIE 14030, Machine Learning from Challenging Data 2026, 140300O

Malarchick, R. et al. (incl. DiMario, C.) "Robust Real-Time UAV Target Tracking with Onboard Vision-Based Yaw Control."

Beyond: Undergraduate Research Journal, Vol. 9, 2025 · Open access

DiMario, C., Bacha, R., & Butka, B. "Combatting Senior Scams Using a Large Language Model-Created Rubric."

ASSE '24, ACM, 2024

Education

M.S. Computer Science

Embry-Riddle Aeronautical University

Expected May 2027

B.S. Computer Science

Embry-Riddle Aeronautical University

May 2026

Beyond the keyboard

  • ERAU Cyber Rodeo 2026: first place, individual
  • ERAU Division II Cross-Country / Track, roughly 30 hours of training weekly
  • FAA Part 107 Remote Pilot Certification
Contact

Get in touch

Open to software engineering roles and internships in embedded, systems, or full-stack. The fastest way to reach me is email.

GitHub

@luccadimario

Résumé · Embedded & Firmware

Download PDF

Firmware on ARM Cortex-M, compiler toolchains, and board-level bring-up.

Résumé · Systems & Networking

Download PDF

Real-time systems, networking, and distributed infrastructure.

Résumé · Full-Stack & Applications

Download PDF

Product engineering across iOS, React, Next.js, and PHP backends.