From Photoshop
to Python. From pixels
to pipelines.

A content engine that thinks like a scientist and writes like a strategist.

When I started at Tisch Scientific, the company had no brand style guide and 10,000+ products that needed unique, technically accurate descriptions. I created the style guide from scratch, presented it to executive leadership, obtained approval — and then built an application to enforce it at scale.

Content Generator 3.0 is a Flask application that takes raw vendor CSV data and produces publish-ready HTML product descriptions. AI helps generate the sentence libraries — but the system I built controls everything else: how those sentences are assembled, rotated for SEO uniqueness, and validated against scientific accuracy rules that prevent errors like claiming nylon filters recover proteins or labeling a 0.45µm filter as sterile.

I also established the grammar rules and marketing voice that the generator enforces. Opening paragraphs use a colleague-professional tone in third person. Product detail bullets continue with direct third-person statements. But in the closing paragraph, the voice shifts to second person for the feature-benefit statements and call to action. The logic: if the reader makes it through all of the content, we've earned the right to speak to them directly. That philosophy is baked into how the system structures every description.