Bernardo – Python, Data Science, Machine learning

A Canadian startup that delivers digital asset analytics powered by AI. Focused on the promising NFT market, it allows customers to measure the fair market value of digital assets accurately.

• Was responsible for technical aspects of the core NFT price prediction engine, including research, development, deployment, monitoring, and maintenance in production
• Developed machine learning models using time series techniques to support EVM and non-EVM compatible blockchains; ERC-1155 and ERC-721 tokens in Ethereum
• Expanded the machine learning system, delivered over a Rest API, to support over 1000 NFT collections and output prices in any token designated by the user, increasing product attractiveness
• Developed a daily automated model performance verification system to update models with low performance in out-of-sample data, successfully safeguarding the user experience
• Integrated leading blockchain data providers via Rest and Graphql APIs using synchronous, paginated, and asynchronous data retrieval processes, with data integrity verified by validation rules

An independent asset management firm in the US market with over $150 MM AUM. The project was a machine learning system to generate market-beating stock portfolios and inform the portfolio manager's decisions. This was a contract project through Virtualmind, working daily side-by-side with the investment portfolio manager.

• Developed a machine learning system for stock portfolio building incorporating innovative financial indicators
• Increased portfolio performance using parameter optimization to achieve up to 30% lift in returns
• Delivered a cloud data product to automate a training and prediction pipeline of US stock market historical data, generating portfolios before market open time
• Developed a model performance monitoring system using a cloud machine learning model logging tool

The product gives managers (and team members) the capabilities they need to measure team engagement and accelerate team performance. The project was to build core product features from scratch as a web app and its integrations in team comms platforms.

• Was responsible for research and development of statistical models leveraging behavior science to measure human factors in remote team management
• Built the foundation of the core feature to distribute hundreds of questionnaires with randomized questions delivered in batches of similar statistical properties
• Collaborated with product research team and leadership to translate value-generating ideas into data models to support new features and business goals
• Developed a data API for end user-facing product dashboards using analytical queries achieving sub-50ms API response time on the 95th percentile with caching

As part of Company's Business Intelligence team, his role was to develop and maintain data processing pipelines and datasets crucial to business management.

• Optimized long-running Apache Spark, Hive, and Presto distributed data cluster jobs and Apache Airflow DAGs, achieving efficiency gains in resource usage and reducing cloud costs
• Maintained datasets updated daily with D+0, D+1, and D+2 SLA with giga and terabyte scales
• Collaborated with product analysts to translate business reporting requirements following data availability and data lineage assessments
• Developed documentation of existing and new datasets and data pipelines, supporting the work of a team of 20 engineers

Brazil's first life insurance startup dedicated to people with chronic conditions. As a founding team member, he could participate in all stages of a growing startup.

• Was responsible for technical research, development, deployment, and maintenance of core life insurance risk scoring system, which calculated the risk score of thousands of recurring payment customers along their subscription and renewal journeys; the system also supported all online price quotes on the platform
• Developed a statistical model from a spreadsheet into a productized system that could output a risk score via a Rest API in collaboration with an international actuarial team
• Developed a machine learning system to classify the quality of online leads. Leads were classified into high, medium and low quality and distributed to the commercial team to maximize team efficiency
• Developed a mathematical model for lead conversion over time, offering deeper insights into commercial team performance and company revenue forecast
• Developed and delivered a complete business intelligence system, including automated and man-in-the-loop daily ETL processes, along with data visualization dashboards for leadership, business development, marketing, and product stakeholders
• Created and led the Analytics team, responsible for risk scoring and business reporting to internal stakeholders and board members
• Supported early-stage lean startup efforts offering data-based evidence and analyses to inform product hypotheses validation and achieve product-market fit