Mobility and location#
How people and vehicles move through cities, from simulation and optimization to privacy-compliant analytics on cellular and WiFi data at scale. Six projects, including earlier work as Lead Data Scientist.
Public health and contact tracing for COVID-19#
Problem · Contact tracing can interrupt the spread of infectious disease when done consistently. A system that combines mobile-phone data with other sources to detect crowded areas and trace contacts needed an independent look at its methodology and algorithms.
What I did · Evaluated the methodology and algorithms of the existing contact-tracing system and delivered a series of recommendations.
Result · The recommendations were implemented, with better performance and more accurate outcomes.
Stack · Elasticsearch, Python, Pandas, data cleansing, trajectory mining.
Optimization in public transportation#
Problem · Bus routes and schedules are adjusted by hand to match demand, fleet and driver availability. For a city network of hundreds of buses and dozens of routes, that relies on individual expertise and leads to suboptimal decisions.
What I did · Designed a solution that estimates origin-destination demand by trip-chaining on entry-only fare-card logs, simulates fleet operations to test scenarios, and finds the best service adjustments against service cost, driver mileage, passenger waiting time and CO2 emissions. Delivered the proposal, statement of work, architecture, jury presentation and a simulation prototype for a European data-incubator bid by a mobility-as-a-service company.
Result · Bid material and prototype delivered. The simulation ran on the real network extracted from OpenStreetMap.
Stack · SUMO, Spark, origin-destination estimation, simulation-aided optimization.
CEO
"We are delighted to work with Michele, who is helping us with the preparation of grant proposals and prototyping."
Machine learning workshop for a mobility-as-a-service provider#
Problem · A company moving into mobility-as-a-service for transport operators needed its sales and management team to understand big data and machine learning well enough to sell and steer it.
What I did · Proposed and led a remote workshop: big data motivations, architectures and integration with legacy systems; machine learning, its implementation challenges and opportunities in mobility; the company's own use cases; and best practices for managing data products.
Result · A shared vocabulary and a set of prioritized use cases for the team.
Stack · Workshop materials, case studies, hands-on examples.
Pre-Sales Analyst
"Clear overview of ML methods with examples applied to our business."
Adaptive traffic signal analysis and optimization#
Problem · Traffic management wants to anticipate congestion and improve the network. Coordinating traffic lights across intersections is one lever to reduce travel time.
What I did · Built an urban traffic simulation from noisy sensor measurements and used reinforcement learning to find the best traffic-light plan, with simulator-in-the-loop optimization.
Result · A working simulation and optimization loop handed over to the research team.
Stack · Flink, Kafka, SUMO, Java, Python, Jupyter, NetworkX, probabilistic modeling.
Principal Research Engineer and Team Lead
"We truly appreciate the effort that Michele put into our collaboration. He is a skilled data scientist that gets things done. We will definitely consider him for other projects in the future."
City-scale people's movement analytics#
Problem · Mobile devices connect to the cellular network constantly, and the logs give a detailed view of how a population moves through a city, useful for purposes such as placing out-of-home advertising, if handled lawfully and at scale.
What I did · Designed and developed a GDPR-compliant analytics service that produces population demographics for urban areas from people's location and movement in cellular network data.
Result · The service ingests billions of records a day in production.
Stack · AWS EMR, S3, Lambda, CloudWatch, HDFS, Parquet, Hadoop, Spark, Scala, Python, Zeppelin, Docker.
Data Science Capability Manager, mobility division of a global industrial group
"Michele is a very talented data scientist with excellent data engineering skills. He contributed several fundamental components of our location intelligence platform. I highly recommend Michele and would love to get the chance to work with him again."
Location intelligence for retail analytics#
Problem · WiFi signals from smartphones locate devices indoors. Retailers use that to find high-traffic areas, evaluate promotions and place products, which needs a pipeline that is both accurate and cheap to run.
What I did · Restructured the WiFi analytics pipeline around Spark jobs, making the data and ML pipelines more efficient.
Result · A double-digit percentage improvement in accuracy and a leaner pipeline.
Stack · PostgreSQL, Airflow, Celery, Cassandra, AWS S3, Python, Docker, JupyterLab, scikit-learn, Pandas, Fabric, Flask.
Lead Data Scientist, later founder of an AI startup
"Michele is a dedicated and thoughtful data scientist. It was enjoyable to see how well he handled tasks and produced excellent, dependable, and repeatable results. Michele is a valuable asset to any company."
Working on something similar?
The quickest way to find out whether I can help is a short call. Prefer email? Write to [email protected].