Portfolio concepts v2 · working prototypes, real WebGL

Krushna Patil
Bengaluru —
fig. 01 — subject
40,000 points

I teach machines
to see.

Open to CV / AI roles
Computer Vision engineer. 4+ years building perception for autonomous driving, robotics and construction drawings — from dataset to 30 fps on the edge.

Your portrait, rebuilt as a 3D point cloud — the way your models see the world. As you scroll, the same 40,000 points dissolve and re-form into each chapter of your work.

The idea

One continuous object carries the whole site. No section dividers, no gimmicks: the point cloud is the story. Move the mouse over it — the points part around the cursor.

On scroll (real site)

  • Self → road (LiDAR rings, cars in the one accent colour)
  • Road → floor plan (Attentive.ai)
  • Plan → node graph (DRIMCO P&ID)
  • Gold medal + contact at the end

Palette & type

  • Near-black, bone white, one signal orange — used only twice
  • Inter Tight (light) + Instrument Serif italic
  • Mono for small technical captions only
KP
input: portrait.jpg · model: kp-v4 · 4+ yrs trained

Krushna
Patil

Computer Vision engineer. I build models that see — roads, robots and construction drawings — and ship them on real hardware.

Krushna Patil
pose: 13 kp · conf 0.97— fps

The homepage runs a model on you. A scan passes over your photo, a pose skeleton and detection box lock on, and the “inference output” becomes the navigation.

The idea

Light, editorial, almost Swiss. It shows taste instead of effects — and the one moment of motion is a joke only CV people fully get, which is exactly who you want to impress.

On scroll (real site)

  • Each log line opens a chapter
  • Chapters use the same overlay language: boxes, masks, keypoints on real visuals
  • 3D used sparingly: one WebGL scene per chapter

Palette & type

  • Warm paper white, ink black, one electric blue
  • Inter Tight semibold, very tight tracking
  • Photo goes from grey to colour once “recognised”
SHEET A-001 · COVER

Krushna
Patil

Computer Vision engineer. Perception for vehicles, robots and the drawings buildings are made from.

4 REV 4 · OPEN TO
CV / AI ROLES
A-001Cover
A-101Attentive.ai — reading plans
A-201Ola Electric — ADAS perception
A-301ICPR-24 — gold, segmentation
A-401DRIMCO — P&ID to graph
A-501Contact

The portfolio as a construction drawing set — the exact kind of document you teach models to read. Every page is a sheet; your career is the revision history.

The idea

Nobody else's ML portfolio looks like this. Ink on paper, a live 3D wireframe of an ego-vehicle with its sensor frustum, title blocks, revision clouds. Calm, precise, very “engineer”.

On scroll (real site)

  • Sheets slide like pages of a plan set
  • Wireframe redraws itself per sheet: car → floor plan → P&ID
  • Detections appear as red-pen markups

Palette & type

  • Drafting paper, ink, one red-pen accent
  • Archivo Condensed + IBM Plex Mono
  • Could also run as a dark “blueprint” variant