Hiren  wanted to be Batman, settled for creative technologist.
I build interactive things — websites for clients, real-time visuals for stages, generative-AI tools for whoever needs them. Self-employed since 2012. Cumulative live audience around 900,000 people across 50+ arena shows in 8 countries. I teach generative animation at the National Institute of Design in Ahmedabad. I'm an open-source-first practitioner — most of the tools I build for myself end up published as MIT-licensed npm packages or public GitHub repos, because collaboration and shared code make the whole field better. The thread connecting all of it is the same: code as a creative material, used in service of a specific person, room, or audience.

Microsites, brand experiences, and case-study sites for studios and cultural institutions. Recent partners: Quicksand, Sangath, Khoj, Pritish Nandy Communications, Obvious, Tryangle42. Webflow, Three.js, GLSL, real-time motion.

14 years of real-time audio-reactive visuals for concert tours, fashion weeks, and gallery installations. Built TouchDesigner pipelines for Arijit Singh, Sunidhi Chauhan, Vishal-Shekhar, Amit Trivedi, Asees Kaur. Introspection at Saatchi + Alserkal.

Generative-AI workflows in ComfyUI, voice-cloning audio plugins, custom LoRA training, GLSL shader practice. And a habit of open-sourcing the tools I build along the way — Refboard, FieldX Waves, Helix, typing-ai.

THE LONGER STORY

Skip if you only wanted the credentials. Keep reading if you want the shape of the practice.

I trained as an engineer at DA-IICT, Gandhinagar (B.Tech, ICT, 2008–2012), but the parts of the curriculum that stuck were the ones where code was a way of making something visible. I founded my own startup — Dotted Yellow Umbrella — in Bangalore right out of college, building multitouch / AR / VR products for real-estate and automotive clients with a four-person team. We shipped both client work and our own hardware-software products for two years, and I learned what running a creative-tech business actually costs.

In 2016 I co-founded Candyman Entertainment with a focus on Indian indie music. Out of that came Fangirl Live, Gujarat's first large-scale fan-experience festival — house gigs in year one, a two-stage two-day festival in year two with 24+ artists, and The Big Squat, a venue we built from the ground up to host an 8-week multi-format festival running independent film, stand-up, spoken word, and music. By the end I'd been Curator, Production Head, and Head of Design & Technology. It taught me how to build the room as well as the show inside it.

From 2019 onward I went officially freelance and started touring as Visual Artist for Arijit Singh Live. Five years, multiple world tours — USA, Canada, Europe, UK, Australia, New Zealand, Malaysia, South Africa, India. TouchDesigner pipelines feeding 8K–12K LED walls, frame-perfect sync with broadcast and lighting, on-site crews of up to 10 technicians, zero-downtime shows. Sunidhi Chauhan's FIFA World Cup opening (Qatar 2022). Lakme Fashion Week as Tech Lead. The Bollywood Music Project as in-house VJ. The All About Music Conference as Tech Lead and Operator. Roughly 300,000 people in rooms I helped paint with light.

I went back to school in 2021 — MA in Interactive Media Art at NYU Tisch — to formalise the parts of my practice I'd been picking up by ear: ml5, p5.js, generative systems, computer-vision interaction. Came out of it with a face-tracking installation, a follow-me Arduino-servo camera, a daily 50-day GLSL practice that became my live-visuals shader pack, and a much sharper sense of what generative work could be when it was deliberate.

Since 2023 I've been Visiting Faculty at the National Institute of Design (NID), Ahmedabad — designing and teaching the Generative Animation elective in p5.js to cohorts of 30 M.Des postgraduate students each semester. Earlier, in 2019, I was technical lead on artist Michelle Poonawalla's Introspection for its international run — a 360° immersive installation shown at Alserkal Avenue, Dubai during Art Week (March 2019) and at the Saatchi Gallery, London (START Art Fair, September 2019).

The current chapter is about generative AI — ComfyUI workflows, custom LoRA training, voice-conversion pipelines, audio-model fine-tuning (VoxCPM, Indic F5), and the open-source tools that fall out of all that work. I post experiments at @metalfinger.xyz on Instagram. The freelance work continues across web, stage, and lab — for clients who want code-as-craft, not template work.

That's the shape. The rest of this page is receipts.

EDUCATION & TEACHING

Year

Institution

What

2009-2013

DA-IICT, Gandhinagar

B.Tech, Information & Communication Technology

2021–2022

NYU Tisch School of the Arts

M.A., Interactive Media Art

2023–2025

National Institute of Design (NID), Ahmedabad

Visiting Faculty — Generative Animation (p5.js, GLSL), 30 M.Des students per semester

2024

Node-Shed Bangalore

Talk: "From Shader to Stage: Real-Time Graphics for Live Music"

AWARDS & PRESS
  • Introspection (2019) — Technical lead on artist Michelle Poonawalla's 360° immersive installation. International run at:
    • Alserkal Avenue, Dubai — March 2019 (Art Week Dubai)
    • Saatchi Gallery, London — September 2019 (START Art Fair)
  • Khoj — The Observatory: Second Site — collaborative WebGL lecture-performance with artist Rohini Devasher and sound designer Legion Seven. Spotlighted by Swiss Arts Council Pro Helvetia.
  • Obvious — case-study microsite (built in partnership with Bengaluru consultancy Obvious) cited in PhonePe's 2024 acquisition note as a brand asset.
OPEN-SOURCE PROJECTS

A core of my practice. Most of what I build for myself ships as MIT-licensed npm packages or public GitHub repos — collaboration and shared code make the whole field better, and I've benefited from that ecosystem enough to want to keep contributing back. Designed to be installed and forked.

Project

What it is

Link

Refboard

Self-hosted, real-time collaborative reference board — PureRef on the web with markdown notes, threaded review, PDF support.

FieldX Waves

npm + CDN library for animated wave / dot-field backgrounds with a visual editor. Designer-first.

Helix

TUI dashboard for monitoring multiple AI coding agents in parallel — htop for your AI agents.

typing-ai

3D ASMR typing scene — wooden puppeteer hands using inverse kinematics to type with you.

WorldBuilder Exporter

Browser-based 3D character customisation tool, the engine behind worldbuilder.in/character-builder.

AddChai Game

2D arcade game (p5.js) that's iframed as the entrance to addchai.com.

AI / ML R&D

Been in the open-source generative-AI ecosystem since the Automatic1111 days; migrated into ComfyUI when it became the better fit for serious workflow building, and have been there since. I treat all of this as one practice: explore an open-source model, train or fine-tune it for a specific use, ship the workflow, contribute back where licensing allows. Image, video, audio, language — same approach across domains. Updated regularly; experiments log lives at @metalfinger.xyz on Instagram.

  • ComfyUI workflows + LoRA training — R&D at Aeos Labs (late 2025): trained three custom character LoRAs — one each on Wan 2.1, Wan 2.2, and LTX 2 — and ran them across eight ComfyUI video pipelines (Wan 2.2 vanilla, InfiniteTalk, HuMo, WanAnimate, latentsync, Wan SVT, Wan Detailer, LTX 2 audio-sync). Output: a recommendation matrix of which pipeline + LoRA wins which job. → /lab/comfyui-lora-experiments
  • Voice-cloning plugin — RVC (Retrieval-based Voice Conversion) wrapped inside a native VST3/AU audio plugin built with JUCE C++ for a film dubbing project. Same model-fine-tuning discipline as the image-LoRA work, applied to voice. → /lab/voice-cloning-plugin
  • Audio model fine-tuning — same practice extended into open-source speech: custom VoxCPM LoRAs trained on ~120 hours of voice data per character (specifics under NDA; checkpoints on private Hugging Face) and F5-TTS fine-tuned for Hindi (also under NDA), where open speech models are structurally under-resourced for Indic languages. Feeds into the dubbing plugin and any future ComfyUI audio workflows.
  • Sound Canvas — SDXL Lightning + ComfyUI API installation for AAM 2025 (commissioned by ChordFather Studio). 800ms img2img loop running on live iPad input. → /lab/sound-canvas
  • ComfyUI multi-modal workflows — depth-to-video networks and ControlNet kinetic typography for volumetric parallax tests in live shows.
  • 50 Shaders / 50 Days — daily GLSL practice; the resulting library is the core of my live-visuals shader pack.
  • Face-Mixing Installation — ml5/p5.js piece that remaps audience faces in real time. Demoed at NYU with 80+ participants.
  • Follow-Me Camera — Arduino-servo webcam controlled by ml5.js face tracking; <100 ms loop. Prototype for retail point-of-purchase displays.

TOOLKIT
  • Real-time graphics — TouchDesigner · Unreal Engine · Unity · GLSL
  • Frontend — Webflow · Three.js · React · p5.js · TypeScript / JavaScript · Vite
  • AI / ML — ComfyUI · SDXL · ControlNet · RVC · LoRA training · F5-TTS / VoxCPM training · ml5.js
  • Programming — JavaScript / TypeScript · Python · C++ · C# · GLSL
  • Design & prototyping — Figma · Adobe CC
  • Audio plugin dev — JUCE (C++)
  • Backend & infra — Node.js · Express · Socket.IO · SQLite · Firebase · MinIO · Docker · Cloudflare Tunnel