Dylan Dalal
Pipeline Development
Many of my projects have focused on optimizing workflows, and I have been specifically employed as a Pipeline Developer at three companies. Choose a specialization for more details.
Saving $8 Million with Vanity A.I.
At MARZ VFX, I contributed to the development of Vanity AI, a production-ready solution designed to handle high volumes of advanced 2D visual effects. The tool specializes in cosmetic enhancements, delivering results up to 300x faster than traditional production methods.
As a Machine Learning Pipeline Intern, I learned the fundamentals of building and distributing AI technology, and how to fit that new technology into an existing company pipeline.
used in 45+ major productions to date
Dylan Dalal Virtual Production
VIRTUAL
PRODUCTION
THE FIRST ACADEMIC VP STAGE
In January 2021, the Florida State University College of Motion Picture Arts launched the Torchlight Center, the campus's hub for teaching film technology. I was their first hire and eventually ascended to the role of Senior Virtual Production Researcher with the responsibility of assembling the first Virtual Production volume ever purchased by a public university.
I researched technologies from many companies and experimented with a multitude before we settled on a Disguise-powered system with RedSpy Stype tracking. On top of wiring and routing data through the volume, I personally contributed to documentation that's been used for class material and trained faculty and staff on how to use the equipment.
It was my introduction to leading a team in a professional setting; to being a part of the planning, testing, and execution of a major project; and to the intricacies of working in a large organization.
PRESERVING HISTORY WITH VR
Before we started the Virtual Production volume, I lead the software development of a VR tour of France’s Notre Dame de Reims Cathedral. FSU partnered with the University of Alabama, who took a 3D scan of the cathedral using Matterport, which we had to clean and optimize before implementing gameplay features.
The tour generated tens of thousands of dollars in funding for the College of Motion Picture Arts for preserving history through technology. It was my introduction to the Software Development Life Cycle and my first major project.
Dylan Dalal Virtual Reality
VIRTUAL
REALITY
Elevating Brands with Elegant Websites
website preview 1
website preview 2
website preview 3
website preview 4
My passion for art and design makes creating dynamic websites a very interesting pastime. In addition to coding the websites, I usually take on the role of graphic designer, video editor, and brand manager. I've worked with a few companies on complete brand redesigns that you can explore below.

Mathematical Property Algorithms

My research at the FSU Department of Psychology centered around removing participant biases from our experiment results by doing in-depth data analysis on respondent datasets. The goal was to determine if and how people use the properties of math when completing math problems.

I never finished the project, which I immensely regret. It was my first long-form software engineering project. Each property was written with a different method, from BET navigation to string processing. It taught me to research the correct tools for the job and how to write a half-decent algorithm.

Short-Form Video Creation Pipeline

Python pipeline that converts Reddit stories into fully-captioned YouTube Shorts: it scrapes posts with PRAW, trims the script with GPT-4o, generates voice-over in ElevenLabs, aligns word-timings via WhisperX, and stitches everything together with FFmpeg into a ready-to-upload 9:16 video.

My goal was to do as little as possible, while uploading as much as possible. This pipeline achieved exactly that.

View repository

C# Webapp Implementation

The resulty of the entirety of my Introduction to C# course, the TaskManager Application provides a full User Interface written in Xamarin.Forms for local user interaction with a WebAPI.

The WebAPI is written using the .NET framework and locally saves inputted tasks to a JSON file for persistent usage of the API.

SVM Classifier Without sklearn

My Machine Learning course at school tasked us with creating a support vector machine from scratch that would draw the most precise line separating two overlapping datasets.

By far the most challenging project I worked on during college, I learned a great deal about supervised machine learning algorithms.

This Website