A large, modular, AI-powered Python automation platform built to orchestrate end-to-end multimodal workflows.
TowleVision is my flagship engineering project. It demonstrates modular Python architecture, applied AI orchestration, workflow automation, and product thinking across a complete pipeline from text intake through narration, captions, planning, image generation, final assembly, social clips, metadata, and publishing-oriented workflows.
The Snow Queen • long-form pipeline showcase
TowleVision_Codex baseline audit — June 2026. TowleVision is a local-first AI media production platform built around structured project creation, narration, captioning, image generation, video assembly, review, repair, metadata preparation, and publishing support.
These numbers come from a June 2026 baseline audit of the active TowleVision_Codex development repository. Generated media, runtime project outputs, caches, dependencies, and virtual environments were excluded from source-code counts. The strict LOC definition is nonblank strict source LOC. The engineering-surface number includes broader source, test, frontend, config, schema, preset, script, and documentation surface.
TowleVision is presented here as a serious flagship engineering project rather than a creative hobby. The outputs matter, but the deeper value is the system behind them: a large Python platform that coordinates dependent stages, integrates multiple AI capabilities, and turns raw input into polished deliverables.
Built as a modular platform with meaningful architectural scope rather than a single-purpose script.
Demonstrates how multiple AI-driven capabilities can be integrated into one coherent production workflow.
Focused on repeatability, usability, accessibility, and finished output quality instead of isolated experiments.
The platform spans a structured workflow from source material through narrated, captioned, image-based video production. It is designed around orchestration, modularity, review, repair, and controlled transformation across many dependent stages.
The transferable value here is not limited to video. TowleVision shows how I approach complex technical systems: define the workflow, modularize the stages, integrate AI where it adds leverage, maintain coherence across the pipeline, and keep the outputs usable and polished.
I maintain a private technical appendix for deeper review, including a polished pipeline diagram, a metrics summary, and a curated repo snapshot.
TowleVision treats AI output as editable production material, not a final one-shot result. The system combines configurable presets, artifact tracking, guarded automation, and human-in-the-loop review so media can be inspected, repaired, and rerun through defined paths.
Projects are organized around source material, presets, generated artifacts, review decisions, repair sessions, caption edits, metadata plans, and downstream rerun paths.
The platform includes Studio Home / Project Workspace, Project Creator and Editor, Script Editor, Project Runner, Video Reviewer, Human Repair GUI, YouTube Upload Assistant, and Music-to-Video Studio.
Dozens of discrete workflow commands are registered and organized so project creation, narration, captioning, image generation, assembly, review, and repair can be run deliberately.
Review and repair layers support images, audio, captions, and final video review, keeping human judgment in the workflow before deliverables are treated as finished.
Upload planning and YouTube publishing support are guarded with dry-run and confirmation safeguards, so publishing-related steps can be reviewed before action.
The June 2026 snapshot shows a working local-first platform direction with broad engineering surface area, while still keeping production claims tied to validated workflows and human review.
These examples show that the platform produces polished, readable, format-aware outputs with attention to caption fidelity, presentation, and multi-format adaptation.
Serif presentation with strong punctuation handling, clean line breaks, and visually integrated captions.
Alternate caption styling for emphasis-heavy moments while preserving readability and polished presentation.
Captioning is treated as a first-class system concern. TowleVision supports multiple polished caption styles while maintaining punctuation fidelity, readable line breaking, and presentation quality across different output contexts.
TowleVision is designed to extend beyond a single output format. This includes vertical short-form presentation, branding, caption handling, and layout choices that feel productized rather than improvised.
These projects are presented as evidence of real system execution, not just as creative samples. Together they show platform depth, output quality, and broader technical range.
The strongest flagship example of TowleVision as a complete system. It highlights end-to-end workflow execution across narration, captioning, story planning, image generation, pacing, and final assembly, while also carrying external recognition as a semifinalist.
A strong example of turning complex source material into structured, accessible output through narration, visuals, pacing, and caption-supported presentation.
Demonstrates flexibility across different literary source material while preserving coherent delivery, visual pacing, and polished finished presentation.
Beyond TowleVision Studio, the broader software portfolio includes applied tools in air-quality reporting, backup and version-control workflows, computer vision, web data extraction, and scientific and technical machine learning.
A Python application for monitoring local air quality through API-backed data retrieval, transformation, and visualization. Included here as evidence of practical, user-facing software development beyond TowleVision.
A computer vision Android app experiment focused on mushroom image recognition and classification-style workflows. Not intended for food-safety, medical, or foraging decisions.
A web data extraction tool built to collect and organize structured information from Kannapedia-style cannabis strain and chemotype resources for research and analysis workflows.
A research-oriented machine learning project exploring cannabis chemotype classification from structured chemical profile data. It is not presented as medically, diagnostically, clinically, or legally validated.
I use TowleVision Insight Lab as the centerpiece of my portfolio for remote Python, applied AI, automation, workflow systems, scientific software, and pharma-adjacent roles. The best way to evaluate my work is through this platform overview, the selected projects above, and my GitHub.
Python automation, applied AI, workflow systems, internal platforms, scientific tooling, technical product development, and pharma-adjacent engineering work.
Based near Seattle, WA and especially interested in remote-friendly opportunities.