Ian Tongs is a machine learning scientist. Normally this is where an encyclopedia would tell you where he was born and what he is known for, in a studiously neutral third person. But I'm writing this myself, so: hi. I build models, argue with data pipelines, and occasionally win.
I work at Comcast, where I build agentic systems and the retrieval layers underneath them, and keep a feature store honest for a few hundred models that would rather not be. Before that I optimized warehouses at Shopify, clustered alarms at Schneider Electric, and modelled electricity markets at MIT. I write about the practice of data science — the parts between the textbook and the job. If you're a fellow practitioner, the writing here is for you[1].
Experience
ComcastPhiladelphia, PA
Senior Machine Learning Scientist, Research, Analytics and Data Science2024 – present
Data Scientist, Enterprise Analytics and Data Science2023 – 2024
- Designed and deployed a multi-agent forecasting system informing C-suite stakeholders on long-term strategic decisions, using supervisor-pattern orchestration with RAG retrieval and structured PDF outputs.
- Built a stakeholder-facing research and synthesis agent, in beta internally, that lets users query company data and external sources directly and generates branded slide decks via a LangGraph StateGraph with conditional retry.
- Built the RAG retrieval layer underpinning both systems — vector store chunking with MMR reranking — to mitigate hallucination and integrate enterprise knowledge sources.
- Maintained end-to-end model deployment through a custom Databricks–GitHub–MLflow pipeline, contributing code to the pipeline itself, and held strict SLAs across a feature store feeding more than 200 production models.
- Designed a prospect-targeting model suite supporting a monthly acquisition campaign of over $10M, including experimentation against incumbent approaches.
Shopify LogisticsRemote / Jersey City, NJ
Data Scientist, Node Optimization Team2022 – 2023
- Developed a proof-of-concept optimization framework for targeted cycle counts, reducing inventory mismatches across the fulfilment network.
- Rebuilt SKU-wise customer inventory reporting across Shopify warehouses, cutting unexplained changes in reported inventory by 90%.
MIT Sloan / Schneider ElectricCambridge, MA
Data Science Collaboration Partner2021
- Analyzed over 30 million device alarm records, developing a two-stage approach using Word2Vec embeddings and adaptive clustering to reduce alarm volume by 65%.
MIT Operations Research CenterBoston, MA
Research Assistant to Professor Georgia Perakis2021 – 2022
- Mined over 100 million pricing observations from US electricity market operators.
- Formulated an optimization approach to electric vehicle and stationary storage purchasing decisions, improving renewable energy usage and customer savings.
Netwealth Asset ManagementMelbourne, Australia
Data Analyst, promoted from Marketing Analyst in 20202017 – 2021
- Investigated causal factors behind customer churn and engagement, and presented intervention strategies to the board of directors.
Education
MIT Sloan School of ManagementCambridge, MA
Master of Business Analytics, Operations Research Center2021 – 2022
Coursework in machine learning under a modern optimization lens, integer optimization, and deep learning. Project work prescribed lifestyle treatments from a 16,000-observation health survey using Optimal Policy Trees.
Monash UniversityMelbourne, Australia
BSc in Mathematical Statistics; BEc in Econometrics2017 – 2020
Monash Community Leadership Scholarship; Dean's Commendation List (Economics, top 5% of cohort); Dean's Award (Science, three years). Mentored underrepresented students through Access Monash.
Recent writing