Develop and enhance digital markeJng module, integraJng voucher and outbound campaign (email/SMS) workflows, supporJng thousands of outlets using Typescript, PostgreSQL, AWS Lambda, AWS Batch, and Step FuncJons

Optimize high-volume backend APIs with database query optimization, pagination, and caching, achieving up to 15x improvement in average response time for several APIs

Train and quantize ResNet-50 and YOLOv7 models with ByteTrack for image classification, object detection and tracking

Build real-time video streaming system (1080p resolution at 30fps) integrating machine learning inference on Rockchip and Qualcomm SoC with hardware-accelerated codecs using GStreamer, multiprocessing, and pipelining in Python

Create Python binding for C++ Snapdragon kernel libraries, streamlining development cycle for on-chip machine learning systems, bringing down inference time by 25 times compared to pure Python implementation

Deliver tools for camera calibration, image orthogonalization, image stitching, distance, and depth estimations with OpenCV, speeding up internal image processing workflow by 55%

Developed web domain credibility ranking models with graph neural networks adopting PyTorch and NetworkX, set up end-to-end ML pipeline, deployed models in production (making prediction for > 1 million domains) using FastAPI

Established data infrastructure (data warehouse, computation, orchestration) on Azure Cloud Services, enhancing data security, quality, and integrity, improving efficiency of data and software teams, and fastening ML life cycle by 40%

Increased efficiency of social media data scraping process by 30% through customization of open-source tools

Analysed tradeoffs in learning algorithms for spiking fully connected and convolutional neural networks

Framed architecture and wrote assembly instructions for spiking neural network hardware accelerators on FPGA

Named on the Honor List of Student Tutors

IT5003 - Data Structure and Algorithms

Conducted classes of ~20 Master students

Reviewed related concepts and techniques, showcased live coding in Python

CS2102 - Database Systems

Facilitated the teaching and discussion of Relational Algebra, Entity-Relationship model, SQL, PostgreSQL, and Database Normalization techniques

Marked and gave feedback on database design and implementation project

CG2111A - Engineering Principles and Practice

Provided consultations during lab sessions, grade lab reports

Related topics: bare-metal programming, communication protocols, Lidar, SLAM, power management

Trained and deployed ML models for scooter parking image classification on AWS SageMaker achieving more than 95% accuracy, detecting bad parking behaviours with 0.98 precision, enhancing user experience, and saving US$200,000 annually

Proposed and coded backend for an in-house annotation platform utilizing Django REST Framework and PostgreSQL, streamlining collaboration between ML and data annotation teams, leading to time savings of 5+ hours per week

Performed experiments with RK1808 neural processing unit (NPU) for on-chip surface detection task on scooters

Designed, developed and tested experimental Physics models

Conducted coaching sessions for students participating in Vietnam National Physics Olympiads