Krushna Patil · Portfolio / Design board v1 — mockups, not the final build

00 — COLOUR OPTIONS (click one to repaint the whole board)

A · Neon Noir

Cyan + magenta on near-black. Classic cyberpunk, the most "neon" option.

#05050A · #00F0FF · #FF2BD6 · #FFE600

B · Acid

Electric lime + violet. Loud, modern, and rarer on portfolios than cyan.

#060806 · #C6FF00 · #7B5CFF · #00FFA3

C · Sensor

Amber + cyan, like an ADAS HUD or LiDAR viewer. Warm, and matches the tones in your photo.

#0A0705 · #FF9F1C · #00E5FF · #FF3B3B

D · Ultraviolet

Hot pink + electric blue on deep purple. Synthwave, the boldest option.

#0B0616 · #FF3D81 · #4D7CFF · #B8FF3D
01 — HERO · desktop 1440
LIDAR · 64-BEAM
SWEEP 212°
OBJECTS 07
FPS 30.0
Open to CV / AI roles · Bengaluru

KRUSHNA
PATIL

Computer Vision engineer teaching machines to see roads, robots and blueprints — and shipping it on real hardware at 30 fps.

See my work →Get in touch
SCROLL ↓
Hero · 3D
  • Real Three.js scene: a road as a LiDAR point cloud, ~60k glowing GPU particles.
  • A radar sweep lights up points; 3D boxes snap onto cars with confidence labels.
  • Mouse moves the camera slightly (parallax); particles scatter near the cursor.
  • Name assembles from particles on load (~1.2 s), then the text fades in.
  • The preview here is a flat 2D sketch of that idea.
02 — CHAPTER: NOW @ ATTENTIVE.AI · scroll section
01 — NOW · ATTENTIVE.AI

Reading blueprints like a machine

Research Engineer 2 on Beam AI — computer vision that reads civil and construction plan sets, finds and measures what's drawn on each sheet, and turns it into estimate-ready quantity takeoffs for contractors.

DetectionSegmentationDrawing understandingOCR
Scroll animation
  • The road particles from the hero fly apart and re-form as a blueprint sheet (morph between point sets).
  • Detection boxes appear one by one as you scroll; a quantity counter ticks up.
  • Section is "pinned" while the animation plays, then releases.
03 — CHAPTER: OLA ELECTRIC · ADAS
02 — OLA ELECTRIC · 2022–25

Perception for 2W & 4W ADAS

L0–L2+ algorithms — adaptive cruise, forward/rear collision warning, pothole & speed-breaker detection — deployed on NVIDIA Jetson and Qualcomm. Plus a robotic-arm pick-and-place system for dark stores.

30
FPS ON EDGE
90%+
ACCURACY
5k+
ORDERS / DAY
Scroll animation
  • Camera drives forward through the point cloud; a car ahead turns red as time-to-collision drops.
  • Stat numbers count up when they enter view.
  • Sub-step: particles form a robot arm picking items (150+ SKUs).
04 — CHAPTER: ICPR-24 GOLD + DRIMCO
GOLDICPR · 2024
03 — AWARDS & RESEARCH

Gold at ICPR-24

Safe segmentation in unstructured traffic — mIoU 0.68, safe-IoU 0.54.

At DRIMCO GmbH: turned P&ID engineering diagrams into graphs — YOLOv11 + OCR at 98% mAP — then let an LLM answer questions about them.

YOLOv11GraphsLLMOCR
Scroll animation
  • A 3D medal spins in with real metal reflections (environment map).
  • Particles form a rotating 3D node graph = the P&ID topology.
  • Mini "chat" bubble: "How many valves feed pump P-101?" → answer.
05 — ABOUT + EXPERIENCE
Krushna Patil
04 — ABOUT

4+ years making models see.

From data collection to edge deployment. M.Tech CSE, IIT Bombay.

  • 2026 — NOW
    Research Engineer 2
    Attentive.ai · Beam AI
  • 2025 — 2026
    CV Research Engineer
    DRIMCO GmbH · P&ID intelligence
  • 2022 — 2025
    Research Engineer
    Ola Electric · ADAS & robotics
  • 2023 — 2024
    Founder Engineer
    Flaunt · AI + backend
Photo treatment
  • Your photo, cropped to a portrait, with a tint in the theme colours.
  • A scan line passes over it and a "human · 0.99" detection box appears — a CV joke that fits your work.
  • Hover: the tint fades and the real colours show.
06 — CONTACT
05 — CONTACT

Let's build
things that see.

kp.iitb.007@gmail.com COPY
LinkedIn ↗Résumé — TODOGitHub — TODO
© 2026 Krushna PatilServed from a laptop in Bengaluru ⚡
TODO list on the site
  • Résumé PDF download button (hidden until you add one).
  • GitHub link (hidden until you have one).
  • "Projects / Lab" section for demos you build in the coming months — designed now, hidden until there's at least one.
07 — MOBILE · 390
SWEEP 212°
FPS 30.0
Open to CV / AI roles

KRUSHNA
PATIL

Teaching machines to see roads, robots and blueprints.

My work →Contact
Mobile
  • Same scene with ~15k particles so it stays smooth on phones.
  • Chapters stack vertically; the 3D sits behind the text.
  • With "reduce motion" on: a still image of the 3D, no animation.
08 — HERO IN ALL FOUR PALETTES (side by side)
A · Neon Noir

KRUSHNA
PATIL

B · Acid

KRUSHNA
PATIL

C · Sensor

KRUSHNA
PATIL

D · Ultraviolet

KRUSHNA
PATIL