Selected work / AI systems & software

Ideas,
made real.

From AI media pipelines to agentic planning tools, RAG-powered assistants, and a live AI app. A look inside the systems we've built and the problems behind them.

A mix of live products, internal tools, and prototypes, all designed and built in-house. Client work is described only with the client's permission.

Product · FFLUCE Scheduler

From media library to scheduled release

Interface / 01View full size ↗ FFLUCE Scheduler content queue with a selected video preview, source assets, and production stages.

The problem

Producing a steady channel of long-form ambient videos means juggling a media library, compositions, renders, titles, descriptions, thumbnails, and upload schedules across separate tools.

The build

FFLUCE Scheduler connects reusable media, composition, AI-assisted titles, descriptions and thumbnails, rendering, and YouTube scheduling in one application, so each video can be followed from library to scheduled release.

Useful when a content operation repeats the same production and scheduling steps for every release.

Status
Working product
My role
Design, engineering, deployment
Built with
FFLUCE rendering · AI-assisted metadata & thumbnails · YouTube scheduling
Read the case study

Client case study · Media automation

From one recording to every format

Interface / 02View full size ↗ Anonymized media production dashboard connecting transcription, edit planning, render variants, and 19 final video outputs.

The problem

A recurring long-form recording needs many publishable versions. Finding excerpts, planning edits, reviewing captions, choosing formats, and keeping final files organized were separate, manual jobs.

The build

A custom application keeps the whole job in one workspace: transcription, AI-assisted hooks and highlights, ranked music excerpts, reviewable captions, batch renders in every format, and an organized delivery folder.

Useful whenever one source has to become many reviewed deliverables.

Status
Working client application (anonymized)
My role
Design and engineering
Built with
Python · FFmpeg · AI-assisted analysis · web review interface
Read the case study

Custom AI software · Shepherd

From scattered context to an approved plan

Interface / 03View full size ↗ Shepherd sample delivery plan showing three tasks arranged in dependency order.

The problem

Project context lives across docs, repositories, tickets, and conversations, and drifts apart as work moves.

The build

Shepherd is an AI project manager that turns repository docs, tickets, and meeting transcripts into structured plans and proposed GitHub work. The team reviews the changes before they are applied.

Useful when people spend more time reconciling the plan than moving it forward.

My role
Design and engineering
Built with
AI planning · GitHub issues · Meeting transcripts
Read the case study

Product · Tavros.ai

One connected system for goals, money, and habits

Interface / 04View full size ↗ Tavros dashboard with a day plan, budget overview, today's transactions and habit check-ins, and a monthly calendar, shown with synthetic demo data.

The problem

Goals, budgets, routines, schedules, and notes usually live in separate apps, so it's hard to see what's working and what's slipping.

The build

Tavros brings goals, money, habits, time, learning, and notes into one system. Describe a goal and it proposes the habits, budgets, study plans, and KPIs to support it, then reports on how it's going.

The same connected-data pattern powers focused business dashboards and internal tools.

Status
Live product
My role
Design, engineering, AWS deployment
Built with
C#/.NET · AWS · PostgreSQL · RAG memory
Read the case study

Consumer AI product · AI Meditation Guide

A live AI app, start to finish

Interface / 05View full size ↗ AI Meditation Guide app screen with a generated session script, voice choices, and background sound settings

The problem

A useful guided session has to fit the person: their intention, the voice they like, and the sound around it.

The build

The app generates a session script from the listener's intention, voices it with synthesized speech, layers background sound, and saves sessions to their account. Built and deployed across iOS and Android with automated releases.

The same building blocks power business tools: generated drafts, natural voices, user accounts, and reliable releases.

Status
Live product
My role
Design, engineering, deployment
Built with
React Native/Expo · LLM & voice APIs · CI/CD
Book a free 20-minute call

New product · Storyboard Director

From musical direction to an editable AI film

Interface / 06View full size ↗ Storyboard Director editing workspace with a video monitor, timeline and editing panels; identifying text is intentionally blurred.

The problem

Creating a complete AI music video involves more than generating individual clips. Shot direction, references, timing, alternate takes and transitions must remain connected as the film changes.

The build

The film plan is the shared structure for planning, generation and editing. Creators review shot direction before rendering, work with saved takes, and follow source generation, transitions and assembly through distinct stages.

Useful wherever AI generates the material but a person directs the result.

Status
New product
Built with
Local & cloud AI video generation · Premiere XML & After Effects JSX export
Read the case study
Photo: Ricky Beron / Unsplash

Start with one workflow

WHAT KEEPS
SLOWING YOU DOWN?

Show us the repeated task, the missing information, or the tool that doesn't fit. In a free 20-minute call, we'll decide whether there's a useful project to take on.

Book a free 20-minute call