amoghbajpai.com

Amogh
Bajpai

Engineering machine perception through the lens of human dramaturgy.

Standby

POV_Calibration

A solo for one face and 478 particles · the audience is the performer

Pixel
Manifold

Everybody is made of pixels. Real-time bitmapping of facial geometry into interactive particles.

Take a breath. Look up. The light is on you.

Stage directions

  • 01. Take the stage to scan facial landmarks.
  • 02. Hold a blink for 3s to charge the color shift.
  • 03. Sustain to 6s to trigger a burst.
  • 04. The intensity meter is your cue light.
listening…

A monologue rehearsed in real time · between you and a stranger that has read your room

Semantic
Stream

Awaiting cue

 

Translating the visual world into high-fidelity narrative descriptions.

Step into frame. The system is watching, trying to put you into words.

Stage directions

  • 01. Begin the monologue. Multimodal inference comes online.
  • 02. Bring objects into frame. Perform an action.
  • 03. The terminal speaks back, every five seconds.
  • 04. Actors. Objects. Latent intent.

An instrument · drop something in, it plays back in 8-bit

Lores

Drop an image, get authentic 8-bit output. Palettes, dithering, no upload.

A browser-only pixel art tool. Game Boy, PICO-8, C64, and seven other palettes. Floyd-Steinberg and Bayer dithering. Everything runs in your browser; nothing is uploaded.

The Dramaturg

I came to engineering through theatre. For years I directed performances, thinking about how bodies move through space, how timing shapes emotion, and how small interactions create meaning on stage.

That way of thinking never left. It simply moved into the systems I build. My discovery of TouchDesigner was the catalyst—it showed me that code could be as plastic and expressive as lighting or movement on a stage.

In theatre, blocking is the choreography of actors within a scene. In software, I think about architecture in a similar way: coordinating agents, managing timing, and shaping how information moves through a system.

Today my work lives at the intersection of machine learning, computer vision, and interactive media. I approach technology as a medium for designing experiences, not just solving problems.

By day I build production biometric systems at Innovatiview. The personal work above is where the rest of me lives.

Tools

Perception

MediaPipe

TensorFlow.js

OpenCV.js

Inference

Gemini · GPT

RAG · LangChain

Visuals

TouchDesigner

GLSL · Three.js

WebGPU

Engine room

Python · FastAPI

TypeScript · Next.js

Postgres

Other code at github.com/amoghgg — face-biometrics-api, biometric-city, fingerprint verification, wildlife detection.

BG FG