Faiz Ansari

Welcome! I write lines on a screen to tell computers what to do - or more colloquially, I’m a software engineer …and my family’s tech support. Two questions, “what can this do?” followed by “what can be achieved with this?”, fuel my curiosity to discover, experiment, and create. As such, my work spans a broad area of development, from full-stack engineering, to reinforcement learning, computer vision, 3D experiences, and game development Luckily, my inner perfectionist ensures that what I build does not come at the expense of quality. I am obsessed with making things beautiful, not just visually, but in efficiency, function, and experience. Say hi at connect@faiz.page

Virtualized Cloud Network

A cloud service for low-latency access to remote computational resources à la AWS' EC2; geared towards high-performance graphics rendering. Built upon an investment of GPUs and PowerEdge servers by Filmtools, a core objective of this service was industry-wide strandardization of CGI and VFX production in fulfillment of the foundational principles of MovieLabs' 2030 Vision Initiative.

Node.js

Express.js

PowerShell

libvirt

Active Directory

Auth0

Watch Demo

Mesh Object Encoder

A utility I created to enable proceedural regeneration of the meshes within 3D environments, such as ones built in Unreal Engine, as part of an effort to establish a remote and collaborative compositing pipeline encapsulated in virutal reality. Written in C++ as an extension for Autodesk Maya, it serializes the vertices, UVs, and textures of selected mesh objects in a scene for export.

C++

Maya SDK

Unreal Engine

RealtimeMeshComponent

Download Demo

Remote Editorial

In response to the COVID-19 pandemic, I created this to facilitate post-production editorial adhering to work-from-home quarantining policies; delivered to clients at Moviola Post. Lets you create private rooms to feature a live hi-res stream of the editor's viewport and the webcams of participants. Self-hosted TURN/STUN servers and dynamic bitrate ensure stablbility and live syncing to the editor's output for all viewers.

Vue.js

JavaScript

NDI Tools

VDO Ninja

Live Demo

The Path Ahead

The Path Ahead is an interactive 3D/VR experience; a dystopic visualization the themes and symbolism of Octavia Butler's Parable of the Sower, created as an academic collaboratory for USC's Ahmanson Lab. I engineered and designed the player control, physics, proceedurally generated terrain, volumetric particle fog, weather effects, crowd simulation AI, and many other core mechanics.

Unity

ProceeduralToolkit

Agile

View Collaboratory / Download

Q-Learning Atari Breakout

Trained a neural network via the Markov decision process to play OpenAI's Atari Breakout. Value and policy iteration executed in TensorFlow, guided by the paper on the problem by Hvass Labs. Found optimal learning-rate and epsilon-probability parameters that averaged a reward of 145.17, a near-perfect winrate, in under 28 training episodes.

Python

Tensorflow

OpenAI Gym

Jupyter

Notebook + Training Logs

NavMesh Pathfinding

Extended upon Brackey's NavMesh Tutorial to create a series of 3D arenas, both static and dynamic, and an agent that traverses them to reach a point on the map selected by the player, follwing an optimal trajectory defined by the A* pathfinding algorithm. Utilizes Unity's NavMesh components to define traversal maps for the agent.

Unity

NevMeshComponents

View Repo / Download

I spend my time practicing calisthenics, tending to my plants, skateboarding, and most recently, script-writing a satire about the human experience. Oh, and obeying the wills of three cats! For most of last year, I volunteered at a communal music-therapy school. My time there involved mentoring students on the drums/guitar as well as being a safe source of emotional guidance/support, according to the unique needs of each student. Here's one of the my last jams at the school, with me on drums.