Hamed Sadeghi

San Francisco CA 650-***-**** advd0z@r.postjobfree.com

Summary

An innovative, metrics-driven, and hands-on Machine Learning Engineer with 11+ years of experience in Machine Learning. Consistently recognized for navigating ambiguity nimbly and admired for being humble and eager to learn, having an agile approach, and concise written and verbal communication skills. Strategic and insightful in ML system design, metrics, and goals feasibility analysis, and providing helpful feedback for data-informed decision-making.

Work Experience

Verily (Google) Life Sciences, South San Francisco, CA, USA

Machine Learning Engineer, January 2022 - Current

●Own On-wrist detection and heart rate estimation projects for the medical watch. Perform full ML cycle A-to-Z design, development, and deployment of algorithms and models in GCP for compliance report

●Design and develop machine learning models and algorithms for On-wrist detection in FW to achieve a 99% sensitivity and greater than 95% specificity

●Design and develop XGBoost models to perform sensor fusion for improving the cloud algorithm specificity (> 98%) compared to the FW implementation while maintaining the FW's high sensitivity

●Design, train, and evaluate deep temporal sensor fusion models to improve the MAPE metric of the heart rate estimation by 4-5 times compared to the previous signal processing-based algorithms

●Manage projects by collaborating with cross-functional teams (science, infra, and FW) and incorporate feedback obtained from meetings, crowd-sourced design doc writing, and software/results documentation

●Document results and software using clear and concise written communication

Amazon.com, Los Angeles, CA, USA

Senior Applied Scientist, May 2020 - January 2022

●Led the science team’s efforts in estimating HVAC (heating and cooling) energy consumption at customers’ homes to help them save energy and reduce their carbon footprint.

●Designed, developed, and evaluated static and temporal prediction models in AWS to estimate daily HVAC energy with a weekly and monthly MAPE error of <30% and documented the results and codes concisely

●Authored roadmaps, scientifically assessed the feasibility of metrics and goals, depicted the timeline of project delivery for stakeholders and relevant project managers from sister teams

Layer 6 AI (TD Bank Subsidiary), Toronto, ON, Canada

Machine Learning Research Scientist, October 2018 - May 2020

●Led the science team to develop predictive models for diabetes onset and complications’ adverse outcomes

●Mentored two interns conducting projects in healthcare using easy-to-understand verbal communication

●Researched and developed deep and classical predictive ML temporal models for diabetes targets using 30 years of 14 Million patients’ administrative health data, achieving AUCs of ~80% with clinical applicability

●Developed graph-based embeddings from the bank multi-task XGBoost models, used it for unsupervised customers clustering, and expedited the development of new use cases by 300%

Huawei Technologies, Montreal, QC, Canada

Machine Learning Research Scientist, April 2018 - October 2018

●Led the science team efforts in significantly improving the subjective and objective quality of the text generation engines using Transformer architecture and deep embedding/adversarial techniques

●Mentored an intern on the text generation project and helped another with the deep model compression

Google, Wellington, Wellington, NZ

Postdoctoral Fellow, February 2017 - February 2018

●Owned multi-channel audio coding project used in Chrome Surround Sound applications. Reduced the storage and communication bandwidth to slightly more than that of a single compressed channel.

●Designed and trained deep adversarial embedding models for 3D audio applications including coding, bandwidth extension, and enhancement.

●Presented results at the Google Chrome conference (internal) in Mountain View, CA in August 2017

Core Competencies

●Industries: Consumer electronics, healthcare, energy, finance

●Teamwork: Collaboration with senior/junior team members and working across teams of stakeholders, and product managers as well as Science, HW/FW, and Infrastructure colleagues

●ML Applications: Prediction, ranking, forecast, clustering, fraud/outlier detection

●ML System Design: development, evaluation, deployment, maintenance, explainability

●Learning Paradigms: Supervised, semi-supervised, self-supervised, unsupervised

●Data Modalities: Natural language, temporal/sequential data, images/videos, graphs

●ML Models: MLP, CNN, RNN (LSTM, GRU, GRU-D, Neural-ODEs), Transformers, tree-based (XGBoost, Random Forests)

Technical Skills

●Cloud Services: GCP, AWS, local HPCs

●Big Data: SQL, Apache Beam, Google Big Query, DataFlow, Flume, DASK

●Distributed Computing: DataFlow, DASK, Google ML Engine

●Deep Learning Platforms: Pytorch, Tensorflow

●Programming Languages: Python, C++, Java

●Machine Learning: Vertex AI, Google Colab, Amazon SageMaker

●Statistics and Signal Processing: Pandas, Sklearn, Numpy, Scipy

●Version Control: Git, Github (https://github.com/hamsade)

●Computer Vision: OpenCV

Education

University of Toronto (UofT), Toronto, Canada

Ph.D., Jan '12-Feb '17, Computer Vision: Image-based localization for mobile and vehicular applications.

Sharif University of Technology, Tehran, Iran

M.Sc., Electrical Engineering, Communications Theory, and Electromagnetics

University of Tehran, Tehran, Iran

B.Sc., Electrical Engineering, Signal Processing, and Communications Theory

Publications

Google Scholar: https://scholar.google.ca/citations?hl=en&user=kbyYu2QAAAAJ

Contact this candidate