Alexander Kimiavi

Led Dedrone’s Computer Vision team, driving ML based video detection, multi sensor fusion, edge deployment, and smart camera control to deliver real time lower airspace target tracking for CUAS and Drone as First Responder missions Led the research and development of Pythagoras 2.0 which improved Dedrone's detection capabilities over Pythagoras 1.0 by more than 30% Boosted throughput of Dedrone edge video tracker by more than 3x by reworking CUDA kernels, optimizing memory access, and vectorizing C++ code Partnered with product managers to translate business goals into clear success criteria and product features

Lead machine learning engineer for Pythagoras 1.0, Dedrone’s flagship object detection model for lower airspace targets, achieving over 24 percent performance improvement compared with the previous model Pioneered small object detection research for ultra low pixel targets, extending effective detection range beyond 2km for small targets and enabling the launch of Dedrone Beyond Lead machine learning engineer for Dedrone Beyond, driving multi-camera detection, tracking, fusion, PTZ control, and edge deployment for Drone as First Responder airspace security Architected and deployed a cloud scale WebRTC streaming pipeline that cut coast to coast video latency from seconds to under 100ms, delivering real time situational awareness for CUAS operators Founded and direct the Data Curation and Perfection (DCAP) program, automating mining and annotation of millions of frames to accelerate model iteration and improve model performance Integrated long range EO/IR camera systems into Dedrone's detection stack, expanding our protected airspace coverage for customers

Develop Deep Learning models using Tensorflow and Python for the use of optimizing hardware utilization of large scale Monte-Carlo simulations across high performance computing (HPC) clusters Work directly with customers to develop Artificial Intelligence and Machine Learning solutions Develop performance analysis tools for streaming based micro-service stacks using Python Increased development efficiency and product delivery speed by utilizing CI/CD into our software development process

Design and develop a micro-service based data analytics platform using Python and Java and orchestrate them using Kubernetes on virtualized hardware Task and mentor interns to support company wide engineering projects and lead technical interviews for potential software oriented new hires

Developed a Health and Status tool using Java, JavaScript, and SQL to monitor the status of components across the DOD Designed a REST API to support the functionality of the Health and Status tool along with the integration of said tool into IDT products

Assisted computer engineering students at Virginia Tech enrolled in micro-controller interfacing with programming and conceptual material taught in the course. Designed projects in C to be used as teaching material for the course.

Developed an automated testing suite for the front and back-end of Localized’s web based application using Python and Django

Designed and developed an automated hardware testing platform for PFP’s proprietary cyber-security hardware using micro controllers programmed in C

Collaborated with the GW Dept. of Biomedical Engineering to research and perform a pilot study on the effects of ”robo-therapy” for children with autism spectrum disorder. Designed and developed prototype robotic toys using Arduino boards and 3D printers

Working extensively on Deep Learning, Deep Reinforcement Learning, and Computer Vision applications. My Master's project is focused on improving existing Deep Reinforcement Learning methods for the use of DNA and Protein sequence alignment. Once completed, the alignment tool will be easily accessible through a Python REST API and Front end web application developed in Flutter.