Location
California, United States
Preferred Work Type
On site
Experience
18+
Years
I. Humayun, a seasoned leader in Machine Learning (ML) and Computer Vision (CV), brings forth over a decade of extensive experience in pioneering the development of Deep Learning (DL) and Generative AI solutions. His expertise spans a wide array of ML applications, including CV, Natural Language Processing (NLP), and Data Science, catering to a diverse global clientele.
Assessments & Skill Tests
Machine Learning
Assess candidates' skills in machine learning, evaluating their ability to develop and implement algorithms, analyze data, and leverage models to solve complex problems and drive insights for informed decision-making.
Score
84/100
Time Taken
120 Mins
Taken On
October, 2022
Deep Learning
Evaluate candidates' expertise in deep learning, assessing their ability to design and implement neural networks, analyze complex data sets, and leverage advanced algorithms for solving intricate problems and driving innovations in artificial intelligence.
Score
85/100
Time Taken
90 Mins
Taken On
July, 2022
Personality Test
C: Analytical and private
C
Taken On
April, 2019
Read more about C
Experience
Director of Machine Learning
PhotoDay
January 2023 - Present
- Directed the ML Team at PhotoDay, leading and managing a team of ML professionals.
- Developed and implemented various ML products to enhance the workflow for volume photography.
- Created advanced algorithms for knockout, retouching, and photo selection for yearbooks.
- Spearheaded projects for culling, spill correction, cropping, and straightening in volume photography.
- Designed and implemented ML models for addressing challenges such as glass glare, teeth whitening, and braces removal.
- Collaborated with cross-functional teams to ensure seamless integration of ML solutions into the photography platform.
- Conducted research and stayed updated with the latest advancements in ML to drive innovation and improve product offerings.
- Mentored and coached team members to enhance their ML skills and foster professional growth.
Chief Scientist & Engineering Manager of Machine Learning
Pano
January 2022 - January 2023
- Developed an ML-based early smoke/fire detection system, specifically targeting wildfires.
- Developed a computer vision-based multi-model system for the early detection of fire & smoke using HD cameras and Geospatial Satellite images.
- Architected and deployed data ingestion pipelines for continuous image acquisition from cameras and satellites, operating 24/7.
- Designed a scalable and compatible architecture to accommodate various types of sensors (such as cameras, satellites, IoT devices), cloud, and edge-based solutions.
- Improved the Time to Detection (TTD) metrics by over 300% by reducing the mean TTD from 7 minutes to approximately 2 minutes, achieving a 100% recall rate on Vegetation fires.
- Designed and implemented a multi-year roadmap for continual enhancement in AI-based smoke detection solutions and infrastructure.
- Built and led a team comprising CV, Geospatial, ML scientists, and DevOps engineers.
Principal Machine Learning Scientist
Mesmer
June 2019 - December 2021
- Architected and developed an AI platform for Robotics Process Automation (RPA) in Mobile App Development.
- Developed DL-based vision models to analyze mobile screens for static and dynamic contents, extracting actionable and non-actionable objects, aiding in crawling the mobile app.
- Designed and developed a DL-based screen_id service to classify and search different screens of a mobile application.
- Designed and developed a DL-based vision object_id service to classify and search different objects within a mobile application.
- Developed an ML-based accessibility rule set.
AI Consultant
Boston Meditech Group
June 2017 - December 2021
- Managed a cross-functional team of 10 ML/CV engineers in 2 locations (Boston and China) to create AI software solutions for a first-generation product. Tripled the number of functional deployed models within the first two years.
- Conceptualized a deep learning breast cancer detection and segmentation framework, achieving 89% recall on the in-house dataset. Had a paper accepted in MICCAI 2020 and a US Patent published in 2022 (US Patent Link).
- Designed a fully supervised model for full breast image malignancy classification. Restructured the problem by identifying global and local-level features, where the latter were extracted using attention mechanisms based on the activation maps. Modernized the BIRADS grading system to be a single process pipeline, reaching 0.75 AUC in early-stage development.
- Strengthened the breast pre-processing pipeline using various image processing techniques and refined the previous design to remove excess muscle parts for CC/MLO images. Reduced the average number of pixels by 10-20% per image.
Lead ML/CV Scientist
Figure Eight
February 2017 - June 2019
- Led computer vision at Silicon Valley’s largest data labeling company, supporting use cases including autonomous vehicles & satellite imaging.
- Was promoted from senior scientist to head of computer vision after the first year.
- Added $5M+ in Annual Recurring Revenue (ARR) with the flagship product, Machine Learning Assisted Video Object Tracking.
- Designed the system to be customizable with a small dataset for each customer, including AV, satellite, medical, retails. (US Patent, US Patent)
- Deployed object detection systems including YOLO, SSD, and Faster R-CNN for a wide range of 2D & 3D objects in RGB and LIDAR. (US Patent)
- Developed an Active Learning technique that progressively accelerated the development of machine learning solutions by allowing the model to query the data they learned from. AL technique with transfer learning and crowdsourcing was evaluated to solve a real-world problem of data selection for training. (AAAI 2019)
- Improved the AVOD framework with a novel class-specific anchoring method for 3D object detection in RGB and LIDAR images. (AIES 2019)
- Designed face recognition and verification systems.
Research Fellow
Harvard Medical School
January 2014 - February 2017
- Received the Komen Postdoctoral Fellowship (3 Years) to develop 3D Computational Pathology (3D-CPath) for analyzing Breast Neoplasia.
- Advanced a novel computational system to discriminate expanded images of early-stage breast neoplastic lesions (Expansion Pathology), significantly improving cancer diagnostics (Nature Biotechnology 2017). A patent application was submitted.
- Developed a crowdsourcing-based image classification technique for the stratification of protein expression in immunohistochemistry images (TMAs) and evaluated it with computational methods and expert labels (Scientific Reports 2017).
- Won the Camelyon 2016 Contest of cancer detection (ISBI 2016). Applied the CNN model to generate tumor prediction heatmaps, which were then utilized to extract geometric features and train a random forest to estimate the cancer probability (Nature Method 2017).
Research Engineer
Next Generation Intelligent Networks Research Center
February 2009 - October 2010
- Worked on a research project titled “Remote Patient Monitoring System” (RPMS). The primary goal of this project was to design a generic remote health care system with a focus on antenatal care, enhancing the services provided by health workers. This project became the winner of the World Summit Youth Award 2009 out of 612 participant projects and runner-up for the P@sha ICT Award 2009 in the research and development category.
- Was responsible for the development of electronic medical record (EMR), web server, mobile and web applications, and an SMS-based notification system between doctors and health workers. Also developed a hybrid decision support system incorporating case-based reasoning and machine learning techniques (Decision Tree C4.5), aiding in diagnosis, medications, and treatments.
- Worked on an offshoot of the RPMS project, focusing on Ultrasound anomalies detection using Image Processing and AI techniques.
Software Engineer
Cogilent Solutions
January 2008 - January 2009
- Conducted requirements analysis, designed, and developed new web portals and affiliates.
- Developed new modules and maintained running products such as BrightSpyre, JobBoards, PersonForce. Used tools like PHP, JavaScript, PostgreSQL, CSS, Ajax, XML, DHTML, Dreamweaver, Joomla.
Software Engineer
RIDOS Software
January 2005 - December 2007
- Clarified software requirements and specifications.
- Designed and developed systems.
Education
Université Joseph Fourier
Doctor of Philosophy (PhD), Computer Science (Medical Imageanalysis)
2011 - 2014
ETH Switzerland
CIMST Multimodal BioMedical Imaging Summer School, BioMedicalImaging
2011 - 2011
Languages
EN
English
Full professional proficiency
UR
Urdu
Native or bilingual proficiency
