// ALEJANDRO_ARTEAGA

> whoami

Alejandro Arteaga

Robotics software engineer. I work on 3D perception, state estimation, and simulation for robots, and on getting that software to run in real time on edge compute.

M.S. Computer Science (Intelligent Robotics), USC, Dec 2026. Open to full-time roles.

On USC's RoboSub team I own software for a 6-DOF underwater vehicle: a ROS 2 DVL driver feeding an EKF 3D-SLAM stack, a 3D RRT motion planner, and containerized ROS nodes orchestrated with K3s on Jetson hardware. I benchmarked LiDAR-inertial odometry on a Jetson AGX Orin, implemented a ray tracer and a software rasterizer in C++, and spent summer 2026 on NVIDIA's Autonomous Vehicles V&V team replacing hand-tuned classifier thresholds with learned decision gates.

3D PERCEPTION & STATE ESTIMATION

  • LiDAR-inertial odometry (DLIO)
  • SLAM
  • EKF sensor fusion (DVL + IMU)
  • Generalized-ICP
  • OpenCV
  • PyTorch
  • Object detection
  • Monocular depth (in progress)

SIMULATION & 3D GRAPHICS

  • Gazebo
  • URDF
  • xacro
  • Ray tracing
  • Rasterization
  • USD
  • OBJ
  • STL
  • Isaac Sim (evaluated)
  • OceanSim (evaluated)

EDGE & GPU COMPUTE

  • Jetson AGX Orin
  • NVIDIA JetPack
  • CUDA
  • TensorRT
  • Triton
  • Profiling
  • C++ multithreading

ROBOTICS SYSTEMS

  • C++
  • Python
  • ROS 2
  • ROS 1
  • Motion planning (RRT)
  • Foxglove
  • Docker
  • K3s Kubernetes
  • Linux

LiDAR-Inertial SLAM on a Jetson AGX Orin

★ FEATURED

EDGE · USC Cyber Physical Systems · Fall 2025

Benchmarked real-time Direct LiDAR-Inertial Odometry (DLIO) on an NVIDIA Jetson AGX Orin, profiling CPU load, core utilization, memory footprint, and computation time across the multi-threaded pipeline and isolating the Generalized-ICP scan-matching step as the dominant bottleneck. Coursework in the same class covered the GPU inference ladder: TensorBoard profiling, Triton, TensorRT, and hand-written CUDA kernels.

  • C++
  • ROS
  • SLAM
  • LiDAR
  • Generalized-ICP
  • Jetson AGX Orin
  • CUDA
  • TensorRT

Ray Tracer & Software Rasterizer in C++

★ FEATURED

GRAPHICS · USC 3D Graphics & Rendering · Spring 2026

Two rendering pipelines in C++ driven by USD, OBJ, and STL scene and mesh files. A recursive ray tracer with sphere, plane, and triangle-mesh intersection, Phong and reflective shading, and BVH acceleration; and a software rasterizer mirroring the OpenGL pipeline, with programmable vertex and fragment shaders, six-plane frustum clipping, perspective-correct interpolation, and z-buffered depth testing.

  • C++
  • Ray Tracing
  • Rasterization
  • BVH
  • Shaders
  • USD
  • OBJ
  • STL

RoboSub AUV Autonomy Stack

★ FEATURED

ROBOTICS · USC RobosubSC · Sep 2024 – present

Software for a 6-DOF autonomous underwater vehicle: a ROS 2 driver for the WaterLinked DVL A50 feeding an EKF-based 3D-SLAM system, a 3D RRT motion planner with dynamic obstacle avoidance and cubic-spline smoothing, and live Foxglove telemetry off the vehicle's Jetson AGX Orin during in-water tests.

  • ROS 2
  • C++
  • EKF
  • SLAM
  • DVL + IMU
  • RRT
  • URDF
  • Foxglove
  • Jetson AGX Orin

Depth Uncertainty Head on DepthAnything

IN PROGRESS

PERCEPTION · USC deep learning course · Fall 2026

DepthAnything predicts per-pixel depth from a single RGB image with no confidence signal. This project trains a confidence head that predicts per-pixel depth uncertainty, using nuScenes driving imagery supervised by its lidar and radar returns. In progress; results to follow.

  • PyTorch
  • Monocular Depth
  • Uncertainty
  • nuScenes
  • LiDAR

IK Solver & RRT Planner for a 6-DOF Arm

ROBOTICS · Fall 2024

Inverse kinematics solver computing joint configurations from Cartesian task-space poses, integrated with an RRT motion planner for collision-free paths. Jacobian pseudo-inverse control held object orientation through a 0.5 m vertical lift.

  • Inverse Kinematics
  • RRT
  • Motion Planning
  • Jacobian Control

Multimodal GenAI Nutrition Assistant

EDGE · Qualcomm Edge AI Developer Hackathon · Jul 2025

Personal nutrition assistant running entirely on-device: Snapdragon inference of the Gemma 3 multimodal model over smartphone food photos and text prompts, with a retrieval-augmented-generation backend on a FAISS vector database for accurate nutritional analysis. Built at the Qualcomm Edge AI Developer Hackathon.

  • Python
  • GenAI
  • RAG
  • FAISS
  • On-Device Inference
  • Edge AI

Software Infrastructure Engineer Intern @ NVIDIA

Autonomous Vehicles — Verification & Validation · Santa Clara, CA

Replaced hand-tuned classifier thresholds with learned decision gates for the Autonomous Vehicles Verification & Validation team, and built the evaluation and experiment tooling around that change.

  • Learned decision gates in place of a hand-authored rules classifier's expert-set thresholds, keeping its input signals and output contract
  • Evaluation harnesses that measure whether the learned classifiers beat the rules baseline on fleet driving data
  • Agent-orchestrated experiment pipeline with human approval gates and CI-enforced operational constraints
  • Python
  • PyTorch
  • scikit-learn
  • Databricks
  • Autonomous Vehicles
  • V&V

Senior Robotics Software Engineer @ USC RobosubSC

Autonomous Underwater Vehicle Design Team · Los Angeles, CA

Leading software for USC's 6-DOF autonomous underwater vehicle — state estimation, motion planning, simulation, and a Jetson-based ROS deployment.

  • ROS 2 driver for the WaterLinked DVL A50 feeding acoustic-IMU data (velocity, pose, altitude) into an EKF-based 3D-SLAM system
  • Custom 3D RRT motion planner with dynamic obstacle avoidance and cubic-spline smoothing for 6-DOF control
  • Led the team's simulation effort as simulation project manager, evaluating NVIDIA Isaac Sim and OceanSim for photorealistic underwater simulation and scoping both against the team's RTX workstations
  • Corrected an 8-thruster misplacement in the vehicle's URDF and xacro descriptions, verified by teleoperation against live ROS topics and Foxglove 3D views
  • Real-time remote telemetry over Foxglove Bridge and a dual-site mesh VPN, used to debug VectorNav IMU and ZED Mini camera data mid-run during bench and in-water tests
  • K3s Kubernetes cluster across a Jetson–Raspberry Pi network orchestrating containerized ROS nodes
  • 99.9% IMU data reliability after diagnosing serial failures (dmesg, lsusb) and hardening udev rules
  • ROS 2
  • C++
  • SLAM
  • EKF
  • Motion Planning
  • URDF
  • Jetson
  • Foxglove
  • K3s

Robotics Software Engineer @ CMU TartanAUV

Autonomous Underwater Vehicle Design Team · Pittsburgh, PA

Simulation, perception, and firmware for Carnegie Mellon's autonomous underwater vehicle on a ROS-based robotics stack.

  • Simulated a depth camera (Microsoft Kinect) in ROS Gazebo, enabling obstacle-avoidance autonomy tests in simulation
  • HSV color-detection pipeline processing RGB feeds at 30 FPS, published to RViz
  • 3,000+ annotated bounding boxes expanding the CenterNet object-detection training set
  • Arduino–Jetson battery-monitor firmware with safety shutdowns preventing power failures
  • C++
  • ROS
  • Gazebo
  • OpenCV
  • Jetson
  • Arduino

ADDITIONAL EXPERIENCE

Software Engineer Intern @ Qualcomm

Modem Software — 5G Access Stratum Automation, Integration & Test · San Diego, CA

Developed automation for 5G Access Stratum modem software integration and test, streamlining validation workflows for the modem software team.

IMPACT
  • Modular Python test-case extractor parsing 5G logs into a searchable SQL database — rapid developer testing against 10,000+ daily test cases
  • Resolved a critical race condition in the multiprocessing cloud upload pipeline, ensuring reliable capture of all test execution data
  • Strict schema enforcement during log parsing surfaced latent testing-framework bugs, improving 5G test-case validity

Software Engineer Intern @ National Instruments

Telemetry Squad · Austin, TX

Contributed to the Telemetry Squad, building and improving telemetry data collection and tooling for NI's software platform.

IMPACT
  • Containerized ETL pipeline prototyped on a distributed Spark cluster for an analytics model
  • 90% ETL runtime reduction by caching Mongo batches in schema-enforced PostgreSQL
  • 30% throughput improvement for large-scale workloads by benchmarking four cluster architectures

Teaching Assistant, 15-281 Artificial Intelligence @ Carnegie Mellon University

School of Computer Science · Pittsburgh, PA

Taught reinforcement learning and sequential decision-making in recitations for three semesters — Markov decision processes, the Bellman equations, and Q-learning.

IMPACT
  • Authored problem sets and exam questions on MDPs, probability, exploration versus exploitation, and Q-learning
  • Edited and maintained the course's Pacman programming assignments

Scientific Computing Research Assistant @ CMU Physics — Neutrino Group

Mellon College of Science · Pittsburgh, PA

Supported neutrino physics research with scientific computing — data pipelines and numerical analysis for experimental physics workloads.

IMPACT
  • 90% speedup of photodetector waveform analysis via C++ multithreading (Michel-electron distributions)
  • Calibrated photosensor energy readings on a heavy-water detector against simulated distributions
  • Quantified magnetic-field effects on photosensor geometry, improving physics-simulator accuracy

University of Southern California

M.S. Computer Science — Intelligent Robotics

USC Viterbi School of Engineering · Los Angeles, CA

Graduating Dec 2026. GPA 3.8/4.0 — University GEM Fellow, Hispanic Scholarship Fund Scholar. Coursework: Robotics, Machine Learning, Cyber Physical Systems, 3D Graphics & Rendering, Extended Reality.

Carnegie Mellon University

B.S. Physics / Astrophysics, Minor in Computer Science

Mellon College of Science · Pittsburgh, PA

Coursework: Artificial Intelligence, Data Structures & Algorithms, Computer Systems, High Performance Computing.