AWS IoT Streaming
An IoT streaming platform on AWS that collects, processes and visualizes real time sensor data, using IoT Core, Kinesis and Lambda for high throughput ingestion.
- AWS IoT Core
- Kinesis
- Lambda
- Python
- DynamoDB
Hi, my name is
Patrick Kozlow.
Solutions architect and growth engineer.
I work with clients and executives to figure out what they actually need, design the system, build it, and stay until the results show up in the numbers. My focus is data platforms, analytics and measurement. I'm currently a Staff Growth Engineer at Headlight, after Instacart Ads and Air Canada's Databricks platform.
About me
Hi, I'm Patrick. I sit between the business and the engineering team: I run the discovery, design the architecture, build what's needed, and make sure people actually use it.
Over the last 8 years I've done that at Air Canada, where I led data engineering for programs that saved $50M+ and returned $10M+ through Databricks; at Instacart, on the Ads data platform behind reporting for enterprise advertisers; and now at Headlight, where I run analytics and measurement for high growth consumer brands.
I'm a Professional Engineer (P.Eng.) with APEGA and hold a Master of Science in Electrical and Computer Engineering from the University of Calgary, where I researched biometric machine learning and computer vision.
Certifications
At a glance
- $50M+
- in operational savings from network wide risk forecasting Air Canada
- $10M+
- in realized ROI delivered through the Databricks platform Air Canada
- $5M
- under budget, delivered on time on the flagship Lakehouse program Air Canada
- $250K+/yr
- in infrastructure savings from the Apache Iceberg migration Instacart
- Solutioning
- Platforms
- Measurement
- Languages
Where I've worked
Staff Growth Engineer @ Headlight
Aug 2026 – Present
- Built Looker automations and client self service tooling, replacing manual weekly reporting with dashboards clients query on their own.
- Own analytics and technical reporting for a portfolio of high growth consumer brands, acting as the technical lead on client calls from discovery through delivery.
- Advise clients and media buyers on web and app conversion tracking, including Meta CAPI, Google Enhanced Conversions and mobile measurement partner setup.
- Design and analyze geo-lift incrementality tests and support marketing mix modeling, turning results into budget recommendations clients act on.
- Build Python ingest pipelines and API integrations on GCP, with dbt models in BigQuery feeding client dashboards, and set engineering standards for the growth team.
Senior Software Engineer, Ads Data Platform @ Instacart
Sept 2024 – Jul 2026
- Partnered with executive sponsors in the advertising organization to design and build a data warehouse on terabyte scale streaming data, delivering real time ROAS, CPI and CTR reporting for 50+ enterprise clients.
- Acted as the escalation point for the enterprise Ads platform, enforcing SLAs across 15+ downstream data products.
- Led the migration of legacy ETL pipelines to Apache Iceberg on AWS, retiring outdated datamarts and saving $250K+ per year in infrastructure costs.
- Built AI assistance tooling with Anthropic APIs and MCP integrations, cutting escalated tickets 40% across 10+ engineers through self service and faster onboarding.
- Delivered a self service Terraform framework for data governance, letting engineers manage production database attributes through standardized CI/CD without platform team involvement.
Data Engineering Lead @ Air Canada
Aug 2022 – Sept 2024
- Ran technical discovery workshops with operations leaders and executives, turning open requirements into architecture roadmaps that met data governance and compliance standards.
- Led the data engineering workstream for network wide operational risk forecasting, identifying 100+ risks at 95% accuracy and generating $50M+ in savings by preventing last minute disruptions.
- Delivered the flagship Databricks Lakehouse program on time and $5M under budget, coordinating multiple engineering teams and vendors, with $10M+ in realized ROI.
- Worked with Databricks and Microsoft on platform cost and performance, cutting compute spend by 20% and deployment timelines by about 30% through CI/CD automation.
- Designed Spark infrastructure processing 50TB of daily flight and passenger data, and moved deep learning risk models from proof of concept to production with MLflow.
Data Engineer @ Nureva Inc.
Feb 2019 – Aug 2022
- Developed ETL workflows with Spark and Airflow to migrate millions of records from an on premises deployment to a cloud architecture.
- Designed a star schema data model for reporting and analysis on over $10M in annual marketing spend.
- Owned on call for production pipelines, cutting mean time to resolution from four hours to under one through reliability improvements and automated alerting.
- Built telemetry pipelines collecting continuous data from thousands of deployed devices, surfacing system failures before customers reported them.
Electronic Design Engineer @ OrbSurgical Ltd.
Oct 2018 – Feb 2019
- Led documentation and prototype development for projects valued over $5 million, aligning technical designs with end user requirements.
- Ran system and component level testing, found the root causes of operational issues, and implemented fixes that improved reliability and performance.
Graduate Assistant, Research and Teaching @ University of Calgary
Aug 2016 – May 2018
- Designed and delivered technical material on Python, SQL, Scala and distributed systems, preparing students for production engineering work at scale.
- Presented research and led technical discussions on biometric applications, data governance and security for academic panels.
Engineering Technical Writer @ Arc Reactions Inc.
2016 – 2017
- Researched and wrote articles on new engineering developments in the energy sector.
- Peer reviewed colleagues' published material and gave feedback.
- Organized interviews and meetings between key people from growing industries.
Things I've built
Gait Trait Analyzer
Contactless biometric gait analysis using computer vision and machine learning. A virtual 3D skeleton extracts gait traits and builds behavioral models in real time for identification.
- Python
- OpenCV
- TensorFlow
- Scikit-learn
- Kinect SDK
Get in touch
Have a data, analytics or measurement problem you'd like a second opinion on, or a role you think I'd be a fit for? Send me a note and I'll get back to you.