Anomaly Detection for Time Series Data using Deep Temporal Convolutional Autoencoder Networks
Developed a Temporal Convolutional Network (TCN) Autoencoder designed for unsupervised anomaly detection and dimensionality reduction in time-series data. The architecture leverages dilated causal convolutions to efficiently capture long-range temporal dependencies, providing a highly parallelizable and stable alternative to traditional RNN/LSTM models. The system encodes sequential data into a compressed latent representation and detects anomalies based on reconstruction error, making it applicable to domains such as predictive maintenance, signal processing, and financial forecasting.
Web Platform for Test Data Normalization
Developed an ETL web platform for normalizing test data across multiple sources. All data sources are tagged and converted using a custom Python pipeline for data cleaning and validation, ensuring consistency and quality.
Real-Time Monitoring for Vehicle Transmission Testing
Designed and implemented a decentralized high-speed monitoring system for lab testing, to allow engineers to visualize and analyze test data in real-time. Leveraged CAN bus fixed-point binary encoding to minimize network overhead, implemented distributed Kafka streaming for real-time data ingestion, achieving <100ms latency from test bench to office while simultaneously uploading terabytes of ASAM MDF files to S3-compatible storage with integrated web preview and analysis capabilities.
Dataset Collection and Asset Management for EV Battery Testing
Developed a comprehensive dataset collection and asset management system for EV battery testing labs using OPC-UA protocol for real-time data acquisition. Built a cloud-native backend with multiple specialized databases: InfluxDB for real-time status monitoring, PostgreSQL for asset management and versioning, S3-compatible storage for test data files, and Neo4j for knowledge graphs and metadata exploration. Integrated with customer cloud infrastructure and enterprise SSO for secure multi-tenant access.
Simulator for Weather Influence on Automotive Lidar and Camera Sensors
Developed a simulator in Unreal Engine 4 for testing and calibration of virtual camera and lidar sensors, enabling autonomous driving validation under various weather conditions.
Lane Departure Warning Algorithm for production vehicles
Developed a lane departure warning algorithm using C++ and faster-than-real-time trajectory prediction models for customer production vehicles.
Creator and Maintainer of Rust Open-Source Library 'map_3d'
Designed and developed map_3d, a lightweight, dependency-free Rust library for high-precision 3D geographic coordinate transformations. The library enables seamless conversion between major coordinate systems serving applications in aerospace, robotics, and navigation simulation.
Creator and Maintainer of Rust Open-Source Library 'fastcurve_3d'
Developed fastcurve_3d, a high-performance Rust library for generating smooth 2D and 3D curves from control points using Chaikin's corner-cutting algorithm. The library is designed for real-time applications in robotics, game development, and computer graphics, providing a computationally efficient method for smoothing trajectories and paths without the overhead of complex spline interpolation.
Path Planning and Control Algorithm for Vehicle Towing Robot
Developed a GPS RTK-based path planning and control algorithm for an autonomous robot that tows cars from the factory to the bay loading area. Utilized ROS for robot integration, Matlab for algorithm prototyping, and Python for sensor processing.
AI Planning and Decision-Making for SAE Level 3+ Autonomous Vehicle
Developed the full path planning and decision-making algorithm for an autonomous vehicle in C++, using AI behavior tree modeling and Lattice planning in a curvilinear reference system.
Control System for SAE level 2+ Autonomous Vehicle
Implemented and tested on test track the planning and control system for an autonomous vehicle in Matlab and Simulink on hard real-time vehicle controller, including adaptive cruise control algorithms, lane steering control and trajectory planning.
Digital Signal Processing and Comparison Tool for Avionic Test Data
Designed and implemented a Matlab tool for helicopter avionics test data that allows comparison and outlier detection within 20ms of misalignment between multiple flight control computers, automating and reporting tests for aircraft certification.