Jacob Danovitch

Experience

Microsoft Search, Assistant and Intelligence

Redmond, WA

Data & Applied Scientist II

July 2023 - Present

Drove experiments improving M365 Copilot grounding data using graph embeddings to increase personalized relevance.
Obtained +0.75 NDCG@1 over baseline ranker & PySpark optimizations saved 100s of compute hours.
Implemented task-agnostic embedding quality metrics, resulting in early identification of model failures.
Promoted to current role in March 2024.

Microsoft Search, Assistant and Intelligence

Redmond, WA / Canada (Remote)

Data & Applied Science Intern

May 2020 - Aug. 2022

Completed 3 internships with Outlook Search Relevance & Graph Intellicence Sciences teams at MSAI.
Used PySpark, PyTorch, & AzureML to build privacy-compliant pipelines analyzing petabytes of data.
Developed novel Graph Neural Networks for search personalization & designed model monitoring metrics.

Microsoft Search, Assistant, and Intelligence

Redmond, WA

Software Engineering Intern

May 2018 - Aug. 2019

Completed 2 internships with Cortana conversational science & engineering excellence teams.
Trained & deployed character-level RNN for intent classification and NER to extract meeting details from Outlook emails.
Named author on patent awarded to parent project: Fenton, Robert Ross et. al, Composing rich content messages. US-20210073293-A1, 2021.
Implemented admin dashboard & API for user feedback application using C#, .NET, and Azure SQL.

Refugee Law Lab - Osgoode School of Law, York University

Toronto, Canada (Remote)

Data Science Lead (Part-time / Volunteer)

Feb. 2021 - Present

Building data platform for legal analytics project from the ground up using open-source software.
Implemented pipelines for scraping, parsing, & analyzing Canadian federal immigration court cases using Prefect, Ray, and Delta Lake.
Distributed version-controlled datasets across research team using LakeFS.
Ensured high availability using Kubernetes (k3s) following GitOps principles (FluxCD / Terraform) and provided secure access over Tailscale VPN.

Canadian Broadcasting Corporation (CBC)

Ottawa, Canada

Data Science Consultant

Sept. 2019 - Jan. 2020

Scraping, cleaning, and aggregating data on health violations in long-term care homes.
Implemented high-performance scraping pipeline to parse semi-structured PDFs from government reports.

Carleton University

Ottawa, Canada

Teaching Assistant

Sept. 2018 - Jan. 2020

Completed teaching assistantships for undergraduate computer science (Web Development, Discrete Math I, Intro to CS I) and graduate-level journalism (Storytelling with Data) courses.
Gave lectures in professors' absence, assisted students in-class, graded evaluations, and held office hours.

Education

Mila - Quebec AI Insititute

Montreal, Canada

MSc Machine Learning | Cum. GPA: 4

Sept. 2020 - April 2023

Researching graph representation learning under supervision of Reihaneh Rabbany at McGill University.
Published at several conferences including NeurIPS 2023, PAKDD 2023, WWW 2021.

Carleton University

Ottawa, Canada

BCS Computer Science, Honors | Minor in Psychology | Cum. GPA: 4

Sept. 2017 - Aug 2020

Graduated August 2020 with high distinction.
Deans' Honour List from 2018-2020.
Completed undergraduate thesis supervised by Professor Majid Komeili.

Selected Publications

Huang, S et al (2024). Temporal graph benchmark for machine learning on temporal graphs. Advances in Neural Information Processing Systems 36.

Huang, S and Danovitch, J and Rabusseau, G and Rabbany, R. (2023). Fast and attributed change detection on dynamic graphs with density of states. Pacific-Asia Conference on Knowledge Discovery and Data Mining.

Pelrine, K^* and Danovitch, J^* and Rabbany, R. (2021). The Surprising Performance of Simple Baselines for Misinformation Detection. Proceedings of The Web Conference 2021.

Fenton, RR et al. (2021). Composing rich content messages.. U.S. patent granted: #US-20210073293-A1.

Pelrine, K^* and Danovitch, J^* and Orozco Camacho, A and Rabbany, R. (2020). Detecting Informative COVID-19 Tweets by Attending over Linked Documents. Proceedings of the 6th Workshop on Noisy User-generated Text.

Danovitch, J. (2020). Linking social media posts to news with siamese transformers. 9th International Conference on Information Technology Convergence and Services (ITCSE 2020), Vancouver, CA.

Danovitch, J. (2019). Trouble with the Curve: Predicting Future MLB Players Using Scouting Reports. 2019 Carnegie Mellon Sports Analytics Conference, Pittsburgh, USA.

Grants & Awards

NSERC CGS-M Research Award

$17500 (declined)

Carleton University

2020

Clarence C. Gibson Scholarship

$3000

Carleton University

2020

Reproducible Research Competition Finalist

$1000

Carnegie Mellon Sports Analytics Conference

2019

I-CUREUS Grant

$2250

Carleton University

2019

E.W.R. Steacie Scholarship

$3000

Carleton University

2019

Lester B. Pearson Scholarship

$3000

Carleton University

2018

President's Scholarship

$3000

Carleton University

2017

Leadership Roles

Carleton University

Ottawa

Recognized Study Group Leader

Sept. 2017 - April. 2018

Co-led study group for introductory CS courses, leading sessions for as many as 50 students at once.
Prepared and presented Jupyter Notebooks to interactive demonstrate course material.
Implemented website for group using Node.js and React, using Firebase for authentication and content delivery.

STEM Fellowship

Virtual

Data Science Team Volunteer

Jun. 2017 - Jan. 2018

Assisted in development & organization of the Big Data Challenge, a competition promoting data science education.
Created workshops for machine learning such as k-means clustering, OLS regression, and stochastic gradient descent.

Carleton Computer Science Society

Ottawa

First Year Representative

Sept. 2017 - April. 2018

Elected to position co-representing 1st year students in the Computer Science program.
Helped design first line of CCSS merchandise and organized sales.