Saad   Hossain
About Experience Publications Resume

Professional Experience

Experience Icon

Jan 2024
|
Aug 2024

Machine Learning Intern

Kolena Inc. | San Francisco, CA

Enhancing the way Machine Learning systems are evaluated.
  • Automated evaluation of Gen-AI by fine-tuning Vision-Language models to model human preference.
  • Improved RAG Generator accuracy by 8% on finance data through improved PDF-parsing & LLM-prompting.
  • Accelerated creation of textual datasets by 15x by building an LLM powered web-application using Streamlit.
  • Enhanced detection of AI model errors for users by building data analysis pipeline identifying edge cases.
Experience Icon

Sep 2022
|
Aug 2024

Machine Learning Developer

Deep Breathe | London, ON

Built AI-powered systems to automate the diagnosis of lung ultrasound scans.
  • Developed AI based screening tool confirming with 96% certainty the presence of fluid in lung ultrasound scans.
  • Achieved 80% accuracy (IoU) in segmenting effusions and consolidations in lung-ultrasound using TensorFlow.
  • Implemented MLOps pipeline using Azure, SQL & OpenCV to process, ingest & version data for FDA approval.
  • Integrated WandB into ML projects for dash-boarding and logging visuals during training and evaluation.
  • Reduced dataset fetching time by 91% through cross-platform deployment of pipeline producing 7.8 million M-Mode images from 140,000 B-Mode videos per version, automated through Unix & Shell scripting.
Experience Icon

Sep 2022
|
Aug 2024

Software Engineering Intern

IntelliSports | Montreal, QC

Developed activity recognition pipelines and modeled user application usage.
  • Boosted cheat-detection accuracy by 30% by feature engineering IMU Sensor data using Scikit-learn & SciPy
  • Optimized inference speed on CPU machine to 1.8x original by leveraging light-weight sklearn models.
  • Elevated mission count by 2.3x by deploying detection pipelines on EC2 for punch, crunch, high-knee, & jump.
  • Identified onboarding flow with +50% retention by conducting decision-tree & SVM analysis on early user data.
  • Built data visualization platform, showcasing user physical activity growth of 300%, employed in pitches that secured $1.4 million in funding using PHP and MySQL.

Research Experience

Experience Icon

Sep 2023
|
Present

Research Assistant

Critical Machine Learning Lab | Waterloo, ON

Investigating ways on how AI models can adapt to new environments.
  • Achieved State of the Art results in Domain Adaptive Semantic Segmentation using Vision-Language models.
Experience Icon

Jan 2022
|
Apr 2023

Research Assistant

Vision and Image Processing Lab | Waterloo, ON

Researched ego-centric pose estimation techniques using spatio-temporal analysis.
  • Outperformed Egocentric HPE State of the Art by 38% via insertion of spatio-temporal data using Transformers.
Experience Icon

Sep 2021
|
Dec 2021

Research Assistant

Social and intelligent Robotics Research Lab | Waterloo, ON

Implemented suite of (IMU) feature extraction algorithms for Human Robot Interaction.
  • Researched and implemented suite of feature extraction algorithms relevant to singla detection for AI.
Experience Icon

May 2021
|
Aug 2021

Research Intern

Human Centered Robotics and Machine Intelligence Lab | Waterloo, ON

Developed and deployed facial and emotion recognition pipeline on humanoid robot.
  • Delivered 99.3% on LFW (Face ID) and 72.7% on FER2013 (Emotion) benchmarks using DLIB and Tensorflow.
  • Designed multi-input Keras model using pybind11 bindings of a C++ toolkit, informed by t-SNE data analysis.
  • Deployed facial identity and emotion recognition pipeline using ROS, Flask and PostgresSQL on the Reem-C robot, highlighted by researchers in International Robotics conferences.