← Portfolio
- My role
- Built and deployed each project with the team it served
- Period
- 2023–2026
- Teams
- HPHS, Regulatory Policy & Economics, ICF sales and executive leadership, Sightline
- Built with
- Django, Next.js, Azure OpenAI, AWS SageMaker, R, Power BI
Needs-assessment assistant
HPHS
The need
An AI assistant for needs assessment, built for HPHS and the client teams it works with.
What I did
Became its technical lead in the second half of 2025 when a teammate left. Supported sprint planning and prioritization, turned business needs into technical specifications, managed the infrastructure as code for its AI Labs and hybrid cloud deployments, and stayed responsive to customer bugs.
The result
Completed the production handoff to HPHS's development team in 2026, with documentation, onboarding, knowledge transfer and a reusable playbook, so the client's own team can run it.
Django 5 · Next.js 16 · Azure OpenAI · PostgreSQL · Pulumi
The need
A public-policy case study needed a Monte Carlo simulation across tens of thousands of scenarios, work that had been bought from an outside subcontractor.
What I did
Rebuilt the simulation workflow in R and Python and ran it on AWS SageMaker, using ScriptProcessor to spread the work and writing results to S3. Also ported the team's geospatial analysis workflow to SageMaker.
The result
Replaced a paid external subcontractor with in-house capability that is cheaper and faster. The Europe and Asia team wanted to reuse the approach.
R · Python · AWS SageMaker · S3
The need
ICF's leaders needed forward-looking estimates of sales and revenue.
What I did
With a colleague, built win-probability, quarter-close and revenue models in R. Built the Azure Data Factory pipeline that feeds them into Power BI, including the connection from the sales database to Azure storage. Presented sales-to-revenue and book-to-bill analysis to the executive leadership team.
The result
The C-suite and executive management use it regularly for financial forecasting.
R · autoXGBoost · Azure Data Factory · Power BI
Account-based marketing models
Sightline
The need
Utilities hold customer data scattered across many systems. ICF's Sightline platform turns it into account-based marketing for commercial and industrial customers.
What I did
Built generative-AI models that predict a business customer's type and parent company. Because the data was sensitive, I compared small local language models (DeBERTa, F-Net, BERT-tiny) inside a secure analytics enclave with frontier models.
The result
My part of the Sightline work that won the Gold Award for B2B Data Excellence at the Association of National Advertisers' 2026 B2 Awards. I was one contributor on a larger team.
Python · DeBERTa · F-Net · BERT-tiny · frontier LLMs