My consultancy. I build every project myself, from first meeting to production.
Built an AI platform for an Atlanta family law practice to manage matters, collect client intake, analyze financial records, and generate court filings.
- Built hybrid retrieval over Georgia statutes, case law, and matter documents, combining keyword and semantic search so AI answers are grounded in cited legal sources and the attorney's case record.
- Built document ingestion and structured extraction for bank statements, tax records, and pay stubs. The system reconciles extracted data with client intake and flags conflicts for attorney review.
- Built client portal flows for intake and financial affidavits, plus generation of divorce filings and parenting plans from case data.
As fractional CTO for a healthcare startup, built a full-stack telehealth platform: patient intake portal, EHR provider UI, and a serverless backend on AWS. FHIR-based architecture on Oystehr with Auth0, appointment scheduling, lab orders, Rx, and automated billing.
- Led HIPAA configuration and BAA negotiations with vendors and hospital partners.
- Built an automated fax parsing pipeline that extracts patient data from incoming faxes and maps it directly to FHIR records, replacing manual data entry.
For an insurance startup, built AI voice agents on Retell AI that stand in for claims staff inside the company's subrogation platform, calling insurance adjusters, navigating phone menus, and verifying demand letters and coverage, replacing hours of manual work per claim.
- Designed an iteration-based versioning system so the team can test and promote new agent versions without touching live calls.
- Built an analytics suite tracking call outcomes, success rates, and campaign coverage.
Designed and built an agentic coding framework that brings AI-assisted development into every phase of the SDLC, from ticket to merged code.
- Skills map to development phases and chain: spec, build, execute, test, pre-review, MR review. Each defines explicit inputs, outputs, and checkpoints.
- Runs inside Claude Code, Cursor, or Codex and pulls live context from Jira, Confluence, and Figma. Works across multiple repos.
- A podcast of a podcast. PodCore distills hour-long episodes into 5-minute audio summaries, plus a written brief with key concepts, notable quotes, and takeaways.
- Subscribe to any show via YouTube URL and get the brief delivered to your inbox the moment a new episode drops.
- Founded and launched Linda, an iOS app that conducts real-time AI interviews and transforms them into polished podcast episodes, published directly to Spotify and Apple Music.
- Named after my mother, inspired by a road trip where I realized I'd never actually recorded her stories. Built for anyone who wants a living audio archive of their own life.
- Users have created thousands of episodes. Average session time of 30 minutes; 45% WAU/MAU for power users.
- Raised angel funding from Factorial Capital and Alex Chung (GIPHY founder).
- First engineer. Built much of the foundational infrastructure: the website, v2 web crawlers, and early content pipelines. Contributed to the API and search engine from the ground up.
- Built foundational systems for a platform that scaled to 2–3 billion daily searches and billions of users worldwide.
- Took on the company's most forward-looking technical problems independently, laying the groundwork for GIPHY's ML org.
- Built and led GIPHY's ML org from the ground up. Grew to a team of 5 ML engineers covering search, recommendations, content intelligence, and generative content.
- Trained and open-sourced GIPHY's celebrity detection model, a custom CNN identifying 2,300+ celebrities at 98% accuracy.
- Built GIPHY2Vec, an in-house NLP model for content recommendations and semantic KNN search.
- Integrated vector search across GIPHY's core search product, achieving 50–200% CTR lift on long-tail queries.
- Built a semantic query clustering microservice that enabled a 5–10x increase in advertiser keyword inventory.
- Fine-tuned generative models for production use: AnimateDiff for video content, CLIP for recommendations, LLMs for search taxonomy.
Part of Atlanta Tech Village's selective advisor program, working with early-stage startups on AI, fundraising, product strategy, team building, and engineering.
- Migrated Alan Lomax's Global Jukebox, a vast ethnomusicology archive, off a local Java server and into a cloud-hosted Django API with a queryable database.
- Uploaded thousands of field recordings and audio files to cloud storage, making them publicly accessible for the first time at scale.
- The project launched to New York Times coverage in July 2017.
Rhizome at the New Museum
2008–2012
- Migrated Rhizome's sprawling PHP codebase to Django, modernizing the platform for one of the world's leading digital art organizations.
- Rebuilt the ArtBase, Rhizome's archive of born-digital artworks, into a custom museum management CMS.
- Collaborated with the Getty to develop a new ontology for describing and classifying digital artworks.
Fractional CTO
I work with your leadership for a set period to review roadmaps and plans, make build versus buy and vendor calls, and guide your developers and vendors.
Focused Production Builds
I take one product or internal tool from scope to launch and hand it off to people who can run it.
Upgrades to Existing Systems
I add AI to products and workflows already in production.
Measurable Results
I set KPIs and baselines, and run tests so your team can see what's working.
If you need a fractional CTO or an AI product built, email nick@stablebasin.com. Based in Atlanta, available for remote and on-site work.