Joseph Marsilla

I believe I have extensive qualifications that satisfy the demands of an ideal applicant for any Data Scientist or AI/Machine Learning/Computer Vision Engineer Role. I currently am enrolled in part-time PhD studies after finishing my MSc in medical biophysics at the University of Toronto, where I worked to harness the power of python based machine learning and computer vision systems to automate tedious radiation therapy planning workflows at Princess Margaret Cancer Center in Toronto, Canada. I have been personally responsible for the development and deployment of translational python based computer vision model(s) in clinical radiation therapy workflows throughout my graduate studies and I believe am more than qualified of taking that same level of translational innovation I use for data science research at my cancer center to accelerate innovation and advanced data engineering at your company or firm.

Proficient in regex, SQL, MongoDB. Advanced skills in developing and deploying python based ML pipelines (PyTorch, Scikit-Learn, numpy, pandas, SITK, ect..).

Describe a situation that challenged your decision-making and problem-solving abilities.

At the beginning of my MSc, we did not have a radiomics group in our lab. I was tasked with the preliminary (hard but doable assignment) of tackling a mammoth head and neck cancer CT imaging dataset with no previous medical image processing experience. Not only did this require building the image and contour extraction protocol from the ground up, but due to the mammoth variability of naming conventions encountered in the DICOM structure sets, I took on the sister challenge of standardizing naming conventions.

Over the period of a month, Single handled, I developed an effective python-based protocol for naming standardization that allowed us to extract all essential radiation therapy targets found in this dataset, which is currently in submission to the TCIA. I went on to use the data that was extracted to train a pytorch-lightning based ecosystem of open-source auto-segmentation models to auto-segment essential regions of interest in the head and neck region. These models are being used in clinical systems at my cancer center.

1. Employment 

• Graduate Student, Dr. Ben Haibe-Kains, UHN (Jan 1, 2019 – present)
• MBP Rotation Student, Dr. Ben Haibe-Kains, UHN (Nov 18, 2018 – Dec 21, 2018)
• MBP Rotation Student, Dr. Anne Martel, Sunnybrook (Oct 16, 2018 – Nov 17, 2018)
• MBP Rotation Student, Dr. Brad Macintosh, Sunnybrook (Sept 2018 – Oct 15, 2019)
• Summer Research Student, Dr. Avi Chakrabartty, UHN (May 2018 – Sep 2018) 

2.

Volunteer

Experience

• Undergraduate Research Student, Dr. Avi Chakrabartty, UHN (Sep ‘17 – April ‘18) 

3.

Scientific

Contributions and Research Outputs

• Conference Talks:
  • CGP-TIP Education Day, Nov 19th, 2020: Overview of Artificial Intelligence: Deep Learning & Radiomics Applications in Cancer Treatment
  • REGEN20, Feb 28, 2020: The Final Frontier for AI in Cancer Research
• Virtual Poster:
  • AACR: Artificial Intelligence, Diagnostics & Imaging - Jan 13-14, 2021: PO-031 Evaluating clinical utility of organs-at-risk segmentation in head and neck cancer by simple open-source 3D CNNs
  • ASTRO 2021, November 01, 2021: Evaluating Clinical Acceptability of organ-at-risk segmentation in head-and-neck cancer using open-source 3D convolutional neural networks
• Pre-prints (MedRxiv):
  • Impact of observer knowledge on AI delineation assessments: Bias in clinical acceptability testing (Co-first Author)
  • Evaluating clinical acceptability of organs-at-risk segmentation In head & neck cancer (HNC) by open-source 3D convolutional neural networks (CNNs) (First Author)
• Publications (Submitted - In Consideration):
  • Development of web-based quality-assurance tool for radiotherapy target delineation for head and neck cancer: quality evaluation of nasopharyngeal carcinoma (submitted to The Green Journal) [Co-first Author]
  • Open-source implementation and clinical acceptability assessment of organ-at-risk auto-segmentation In head & neck cancers (Submitted to eBioMedicine) [First-Author]
• Open-Source Repositories:
  • PyTorch Lightning OAR Segmentation (sole-contributor) - Boilerplate for PyTorch Lightning based medical-image segmentation pipeline for to enable automated segmentation of organs-at-risk (in [but not limited to] head and neck cancer) by using Simple Open-Source 3D CNNS
  • QUANNOTATE (collaborator, original repository used in QANPCRT coded in python-flask) ; QANPCRT: Quality Assurance of NasoPharyngeal Clinical Radiation Targets (collaborator - image processing, construction of web-based QA tool)
  • Pygempick – Open-Source python module to enable batch gold particle picking, filtering, processing & statistical modelling of TEM images produced by IGEM (immunogold electron microscopy).
  • Pygempick-flask - This is the pyGemPick web application built on top of Flask python framework allows researchers to count and process statistics of immunogold particle EM image sets automatically.

4. Honours

• Sep 2019-2020: Princess Margaret Hospital Foundation Graduate Fellowship in Cancer Research – MBP Excellence Award
• 2015-2017: Dean’s List Scholar, University of Toronto
• Nov 2014: St Michael's College/c L Burton Trust Fund In-Course Scholarship
• Oct 2013: President's Entrance Scholarships, University of Toronto
• April 2013: Ontario Hockey Federation Scholarship

In-Progress:

• PhD (part-time), Faculty: Medical Biophysics, University of Toronto, April 2026 (Supervisor: Dr. Ben Haibe-Kains)

• MSc, Faculty: Medical Biophysics, University of Toronto, April 2022 (Supervisor: Dr. Ben Haibe-Kains)
• Hon. B.Sc. Biological Physics Specialist [Biochemistry Stream] with Math Minor.

  Faculty: Arts & Science, University of Toronto. Date: June 2018 (with High Distinction)