Developed a ROS2 autonomous navigation pipeline using Python, OpenCV, PyTorch, and LEDNet for real-time trail segmentation, combining YOLOv7, lidar, and camera fusion for pedestrian-aware robotic navigation.
Engineered a 3D vision pipeline using Python, OpenCV, and YOLOv3 on KITTI stereo images, fusing depth maps with 2D detections to segment vehicles and estimate positions and distances.
Built a Python pipeline using OpenCV, HSV segmentation, and K Means clustering to detect targets from drone imagery and transform them into global coordinates via calibrated camera models.
Created a 3D point cloud alignment system in Python using the ICP algorithm, nearest-neighbour search, and SVD to estimate optimal transformations between source and target scans.
Designed a MATLAB/Simulink quadrotor control system using state-space modeling, kinematics, and cascaded PID control to simulate UAV flight dynamics, stability, and path tracking performance.
Developed autonomous drone racing capabilities using Python, RRT*, and PyBullet, combining trajectory optimization, polynomial path fitting, and sim-to-real deployment on a Crazyflie quadrotor.
Developed a Python visual odometry system using OpenCV and NumPy to estimate vehicle pose and localization from KITTI stereo images, leveraging RANSAC, point cloud alignment, and SVD.
Implemented an Extended Kalman Filter for 2D robot localization, fusing wheel odometry and laser range data to estimate pose under noisy, partially observable environments with uncertain measurements.
Led software development for a ROS-based autonomous robot using Python, C++, LiDAR, Hector SLAM, and AMCL for mapping, localization, and navigation with real-time obstacle avoidance.
Developed autonomous navigation algorithms for Turtlebot using ROS, Python, C++, and Gazebo, combining heuristic planning and sensor feedback to optimize exploration paths in unknown environments.
Implemented optimal trajectory planning for a 2-DOF robotic manipulator using CasADi and IPopt, applying direct optimal control to optimize motion for energy, time, and torque objectives.
Compared sampling-based motion planners in Python by applying RRT and RRT* to a Dubins vehicle, demonstrating improved path optimality through rewiring while respecting vehicle constraints.
Developed a data science pipeline using Azure ML, SQL, and machine learning models to classify human activity from sensor data, leveraging AutoML to achieve 94.1% accuracy with ensemble learning.
Automated a Google Cloud data pipeline using Python, SQL, Apps Script, Cloud Functions, and BigQuery to ingest, clean, store, and analyze environmental sensor data from Antarctica.
Analyzed petabytes of vehicle telemetry data using SQL and BigQuery to map global modem performance, integrating Mapbox visualizations to uncover geospatial trends in cellular connectivity.
Built a scalable geospatial analytics pipeline using SQL, BigQuery, and DBSCAN to identify public gas stations from millions of vehicle fuel events while filtering false positives.
Trained and refined a multi-class CNN classifier using validation tracking, Adam optimization, learning rate scheduling, and regularization techniques to improve animal image recognition performance.
Implemented reinforcement learning with a neural network policy using the REINFORCE algorithm to solve OpenAI Gym’s CartPole-v1 task through Monte Carlo returns and gradient-based optimization.
Implemented deep learning-based semantic segmentation in PyTorch to classify outdoor trails, benchmarking multiple CNN architectures and optimizing LEDNet for accurate real-time robotic perception.
Built a custom Transformer architecture for machine translation using deep learning and NLP techniques, implementing attention mechanisms, positional encoding, and decoding strategies for language generation.
Prototyped an embedded control system using Arduino Mega, integrating PWM actuators, UART/I2C communication, sensors, and user interfaces to automate a custom drink-mixing machine.
Prototyped an Arduino-based embedded security system for temperature-controlled transport, integrating sensors, servo actuators, keypad access, and Fusion360 CAD design during a 24-hour design challenge.
Designed an autonomous maze-navigation robot using Raspberry Pi and Arduino microcontrollers, integrating LiDAR, sensor fusion, and real-time communication via USB, I2C, and SPI protocols.
Created an embedded human-robot interaction system where a sensor-equipped glove controlled a robotic vehicle via gestures, combining Arduino, Raspberry Pi, ROS, and wireless communication.
Investigated hardware reliability in embedded telematics devices through eMMC wear analysis, using diagnostic data to assess memory degradation, wear leveling, and potential failure modes.
Designed a custom 2-layer analog PCB in Eagle for an oscilloscope system, implementing signal conditioning, op-amp amplification, filtering, protection circuits, and power conversion.
Integrated a 288V lithium-ion battery system for a high-speed EV, performing cell testing, thermal validation, and electrical assembly of PCBs, converters, sensors, and power electronics.
Simulated a solar thermal power tower in Energy3D, analyzing heliostat layouts and receiver configurations to evaluate design factors affecting renewable energy performance.
Created a SolidWorks CAD model of a MacBook Pro, translating manufacturer specifications into a complete assembly with detailed components, textures, and design features.
Redesigned an outdoor rowing machine using MATLAB-based dynamics analysis and a parallelogram linkage mechanism to stabilize force output, optimize motion, and improve user performance.
Created a CNC milling machine design in SolidWorks, combining mechanical design, component sourcing, and manufacturing documentation to develop a reproducible machining system.
Applied computational and experimental methods to assess aircraft stringer deformation, using ANSYS FEA, mesh convergence, tensile testing, and strain analysis to study structural failure.