Case Studies

The Problem

Important portfolio and product information lived across financial and project/portfolio management systems, which meant employees often had to build reports manually just to understand project status or take action.

What I Owned

• Owned agent integrations with internal systems and helped design and implement the main agent experience.
• Built core agent infrastructure, retrieval workflows, and supporting data models.
• Worked across data modeling, reliability, and production-readiness concerns rather than stopping at prototype-level functionality.

What I Built

• Used Snowflake Cortex, dbt, and Airflow to connect governed enterprise data with agent workflows that could answer questions and support operational workflows.
• Optimized queries and recovery behavior so the main agent could handle prompts touching years of data while remaining fast and resilient to unexpected inputs.
• Tuned guardrails for governed updates to balance safety, accuracy, and responsiveness without making the experience unusably slow.

The Problem

Forecasting workflows are high-stakes and organization-wide, and they become painful quickly when business logic changes are slow to intake, hard to test, or messy to deploy.

What I Owned

• Worked as a data engineer on the pipeline and surrounding engineering workflow for the forecasting product.
• Contributed directly to the dbt-based transformation layer and the delivery discipline around testing and deploying changes through GitHub.
• Helped improve how new business logic was ingested, structured, and turned into implementation work.

What I Built

• Used dbt and orchestration workflows to support a product where stakeholders maintained planning inputs and projected spend during key planning cycles.
• Relied heavily on GitHub CI/CD, branch protection, testing in pull requests, and structured deployment workflows to keep changes safe and reviewable.
• Used automation and subagents to accelerate business-logic intake, implementation, and code review workflows once the process was mature enough to standardize.