All projects. Professional or personal, live or prototype.

BUSINESS PROBLEM

A repeat client was handling sales operations, data analysis, and prospecting by hand across spreadsheets and disconnected tools.

WHAT I BUILT

Two separate engagements for the same client: a field sales tool and a data analysis workflow. Each was scoped to one specific bottleneck rather than a platform rebuild, and each was maintained after delivery.

RESULT

Two engagements with the same client, including a tool in daily use for several months and follow-on work currently being scoped.

BUSINESS PROBLEM

A women's hiking group was coordinating monthly hikes, RSVPs, and attendance by hand across text threads and email.

WHAT I BUILT

A member-facing site with passwordless sign-in, a published season schedule, self-service RSVP that members can change, self-marked attendance history, and automated reminder emails. Admins control roster visibility from a settings toggle rather than needing a code change.

RESULT

Live with 30 registered members and the full season published for RSVP.

BUSINESS PROBLEM

A private network of wine enthusiasts wanted a place to rate wines already tried and track the ones still on the list, something more personal than a general-purpose rating app.

WHAT I BUILT

A wine-logging app, invite-only by design, with an uncapped invite chain doing the rest of the growth: a small seed group gets a code, and from there the network grows through invites those users choose to send, with no cap set from the top.

RESULT

Live and invite-only. Three seed invite codes went out at launch, and each holder is free to invite as many people as they want, so the initial cohort is a seed for however far the network wants to grow on its own.

BUSINESS PROBLEM

Finance teams manually retype information from incoming invoices into spreadsheets and accounting systems.

WHAT I BUILT

An automated pipeline that reads invoice PDFs, extracts the key fields, checks whether the extraction is confident enough, routes uncertain invoices to a person for review, and writes approved records into a spreadsheet.

RESULT

Demonstrates a production-shaped document workflow with confidence gating and human review. Built and tested against sample invoice PDFs, not live production invoices.

BUSINESS PROBLEM

Notes, sources, and research accumulate faster than anyone can organize them, and asking an AI about your own material usually means it can't tell a solid source from a weak one.

WHAT I BUILT

A knowledge system with a conversational retrieval interface, a 3D map of how the knowledge connects, and a trust-scoring layer that rates how well each retrieved source can actually support the claim it's being used for.

RESULT

In daily personal use. Development surfaced and fixed a real defect: a source-deduplication bug that let one document be counted several times and inflate a claim's confidence score. The first assertion-based test suite in the project caught it.

BUSINESS PROBLEM

Support teams spend most of their time answering the same routine questions, while genuinely difficult cases wait in the same queue.

WHAT I BUILT

A support workflow design where AI answers routine requests from a knowledge base, checks its own confidence before responding, escalates uncertain or sensitive cases to a person, confirms satisfaction, and raises priority when a customer isn't satisfied.

RESULT

A worked design for human-in-the-loop support: routing rules, confidence thresholds, escalation of uncertain or sensitive cases, and satisfaction follow-up. This is a designed workflow, not a deployed product.

BUSINESS PROBLEM

A test of whether careful prompt design and published analytics frameworks can beat naive prediction on a hard forecasting problem.

WHAT I BUILT

An NCAA tournament prediction system built on published basketball analytics frameworks, run live through the full 2026 tournament with bracket tracking.

RESULT

73% game accuracy across 63 games, against a roughly 75% benchmark from established models. One tournament is a small sample and the approach carries real overfitting risk. The case study explains both rather than treating the number as settled.

BUSINESS PROBLEM

Form responses pile up and nobody reads them until someone needs a summary for a meeting.

WHAT I BUILT

A workflow triggered by form submissions that runs each response through sentiment analysis, logs the categorized result to a spreadsheet, and sends one aggregated summary email across all submissions rather than an alert per response.

RESULT

Demonstrates batched analysis and reporting over a stream of form responses. Built and tested against sample submissions.

BUSINESS PROBLEM

A researcher needs answers grounded in one specific source, not a model's general recollection of the topic.

WHAT I BUILT

A chatbot constrained to a single academic source document by design, answering only from that source and formatting every response to Chicago Manual of Style citation standards.

RESULT

Demonstrates constrained retrieval with enforced citation formatting. Built against one source document.

BUSINESS PROBLEM

Dispatchers find out a route is running late when the driver is already late.

WHAT I BUILT

A workflow that reads the day's routes from a spreadsheet, runs two sequential analysis passes over the same data to assess cumulative delay risk, and sends either a delay alert or an all-clear summary based on configured thresholds.

RESULT

Demonstrates threshold-based operational alerting over route data. Built against sample routes.

BUSINESS PROBLEM

Operations managers rebuild the same weather-versus-staffing judgement call every morning.

WHAT I BUILT

A workflow that pulls the daily forecast and branches into one of two analysis prompts depending on whether conditions are severe or standard, delivering a tailored staffing and operations recommendation each morning.

RESULT

Demonstrates conditional branching into distinct prompts based on external data. Built and tested against live forecast data.

BUSINESS PROBLEM

A field sales rep was managing daily routes, visit notes, and dozens of retail accounts, meaning individual stores, across a phone, a notebook, and memory.

WHAT I BUILT

A mobile-first field sales tool for managing the daily route, logging visits as they happen, and seeing active accounts on a map instead of a list.

RESULT

In daily use by one field sales rep covering 30+ retail store locations.