Habib Moukalled
Senior Software Engineer | On-Device ML Inference & Sensor-to-Model Data Paths
Hire this person
Send a job offer directly to this candidate
About
Systems engineer who makes trained models run fast and reliably on real hardware. For the past five years I have built the inference pipeline and sensor-to-model data path for real-time biometric authentication on Android devices at Google: acquiring camera and sensor data, preparing it for an on-device TPU inference engine, and measuring end-to-end latency. Before that, 10+ years of high-performance computing in C/C++, including GPU kernels (CUDA, OpenCL, Metal), SIMD math libraries, and low-latency video pipelines.
I focus on everything around the model: getting data into it quickly, running it efficiently, and proving the timing with automated tests.
Experience
Senior Software Engineer - Google (contract via AutoRoboto, LLC) - Remote
(2021-03)
- Built the sensor-to-model data path for real-time biometric authentication: camera and sensor acquisition feeding an on-device TPU inference engine across multiple Android products, profiling the pipeline to find slow stages and fixing them so it stays within the product's latency quality-of-service bar.
- Designed a cross-product hardware abstraction layer (HAL) so one inference pipeline runs on different device hardware without per-product rewrites.
- Built sensor and image ingestion that supplies training and evaluation data to ML pipelines, working with modeling, hardware and product teams.
- Developed automated tests that simulate real user interaction and measure end-to-end inference latency, [running across X devices / catching X regressions].
- Built the system that rolls ML feature changes out to users as controlled experiments, comparing treatment and control behavior.
- C/C++, Python, Android, Linux, DSP, on-device TPU
Senior Software Engineer - 2K Games (contract via Leviathan Games) - Remote
(2020-10 - 2021-03)
- Built a cross-platform, SIMD-optimized linear algebra library (x86 AVX/SSE intrinsics) powering the game engine on PC, PlayStation and Xbox.
- Designed math APIs used by AI, camera and physics teams.
- C/C++, Python, Jenkins
Senior SW & HW Engineering Support Operations Manager - The Coca-Cola Company (Freestyle) - Atlanta, GA
(2019-06 - 2020-10)
- Led triage and resolution of escalated software and electronics issues across a fleet of 50,000+ network-connected beverage dispensers.
- Built Python tools to analyze terabytes of SQL and log data, pinpointing root causes of field failures.
- Mentored junior data scientists, software engineers and senior technicians; ran cross-functional investigation meetings.
- Lead inventor on patent for remote beverage selection (WO/2022/082102).
- Python, C/C++, Java, SQL, Splunk, S3, IoT
Senior Software Engineer - AstroHQ - Remote
(2017-04 - 2019-05)
- Architected and maintained versions 2 and 3 of the LIQUID low-latency video engine, moving frames from capture to display with minimal delay.
- Wrote parallel GPU algorithms in Metal and OpenCL for the video pipeline.
- Built a rate-control algorithm that adapts throughput to measured network conditions.
- Built Python tools to model the video pipeline and analyze algorithm performance.
- C/C++, Objective-C/C++, Metal, OpenCL, OpenCV, macOS, iOS
Senior Software Engineer - Euclid Discoveries - Remote
(2015-10 - 2016-10)
- Built a Python test harness and parallel test framework for automated QA of a video encoder library.
- Built utilities that detect and characterize encoding performance anomalies; produced custom FFmpeg, H.264 and H.265 encoding environments.
- Deployed test infrastructure on AWS in Docker containers.
- C/C++, Python, AWS, Docker, Boost
Owner - HabiSoft, LLC
(2015-10 - 2019-06)
Consulting on computer vision, automation and algorithm performance optimization using CUDA, OpenCL and OpenGL.
Software Engineer - MIT Lincoln Laboratory (contract) - Lexington, MA
(2013-07 - 2015-10)
- Built hard real-time image and signal processing for large data sets, with parallel CPU and GPU (CUDA) algorithms.
- Developed computational models and data visualization for image and signal data.
- C/C++, Python, CUDA, OpenGL, OpenCV
Software Engineer - Applied Quantum Technologies - Durham, NC
(2012 - 2013)
- Signal processing on data sets of 1M+ points for government-funded projects.
Software Developer - University of South Carolina - Columbia, SC
(2006 - 2012)
- Computer vision algorithms for high-speed video analysis.
Education
M.S. - Computer Science - University of South Carolina (2009)
B.S. - Computer Science - University of South Carolina (2007)
