Work

Shipped. Measured. Still running.

Build stories in plain language — the problem, what I built, and the status. Where you’d expect inflated metrics, you find honesty instead: numbers appear here the moment they are real.

01

3D Streaming TV Platform

Immersive web · streaming

shipped
Shipped WebGL Platform
60 FPS CAPTURE
3D Streaming Television Interface3D Interactive Preview · Tilt with Mouse

The problem

A TV channel's site was a flat list of shows. Viewers scrolled past it the way they skip ads — nothing about it felt like television, and nothing on it turned a visitor into a returning viewer.

Stack

  • React
  • Three.js
  • 3D pipeline
  • Performance engineering

The build

  • Designed and built a fully immersive 3D browsing experience — depth, camera motion, and spatial navigation instead of static grids.
  • Engineered it to stay fast under real streaming traffic: progressive 3D loading, poster-frame first content, transform-only animation.
  • Wired the content pipeline so programming updates don't need a developer.

Status & outcome

Shipped as a production site for the channel. The build demonstrates the exact combination this studio sells: cinematic 3D that still performs. Metrics section appears once real channel numbers are supplied.

Engineering Teardown · Impact Comparison

Traditional 2D Flat Web vs. 3D Spatial Retention

● 3D Spatial Storytelling (Mitiku)
60 FPS WebGL Spatial Depth

Unbroken camera choreography, dynamic PBR lighting, and spatial audio cues.

Conversion4.2% (+300%)
Avg Dwell2m 54s (3.8x)
Bounce Rate24% (-44%)
○ Traditional 2D Flat Web
✕ Static 2D Card Grid

Flat image cards with generic scrollbar. Viewers skim past like a standard banner ad.

Conversion1.4% (Industry Avg)
Avg Dwell42 seconds
Bounce Rate68%
⇄

Drag slider left & right to evaluate the conversion & dwell time difference.

02

Restaurant Management System

Full-stack system · operations

shipped

The problem

Running orders, tables, and staff coordination through word-of-mouth and paper creates slow service, lost orders, and end-of-day guessing.

Stack

  • Full-stack
  • Real-time operations
  • Data modeling

The build

  • Built a full-stack system covering ordering, table management, and operational reporting.
  • Designed the flows around the actual shift rhythm of a restaurant — speed first, everything else second.
  • Structured data so every order becomes queryable history instead of a paper trail.

Status & outcome

In daily operation. It is the proof behind the automation service: I don't just build interfaces — I build the systems a business runs on. Metrics section appears once real operational numbers are supplied.

System Architecture · High-Concurrency Breakdown

Real-Time Order Ingestion Under Heavy Shift Load

Throughput14 orders / min
P99 Dispatch Latency1.8 ms
Packet Drop Rate0.00% (Zero loss)
Connected KDS Displays6 Station Queues
TIER 01HTTPS
Client Ingestion Edge

QR Mobile Ordering & POS Terminals

Zero input latency○
TIER 02WSS
Real-Time Event Broker

Pub/Sub High-Concurrency Gateway

Sub-4ms dispatch●
TIER 03Station
KDS State Machine

Kitchen Display Routing Matrix

Deterministic FIFO & rush prioritization○
TIER 04ACID
ACID Ledger & Inventory

PostgreSQL & Reconciliation

100% financial balance & inventory sync○
[Tier 02] Real-Time Event Broker· Pub/Sub High-Concurrency Gateway
Protocol: WSS / Binary JSON-RPC
Engineering Role

De-duplicates incoming tickets, validates idempotency keys, and broadcasts instantly to prep displays.

Production Guarantee

Sub-4ms dispatch, zero dropped packets

03

Agentic Storefront

3D + AI agent · e-commerce

● in progress
Flagship Multi-Agent Platform
60 FPS CAPTURE
Agentic Storefront & 3D Catalog3D Interactive Preview · Tilt with Mouse

The problem

E-commerce loses money in three quiet places: unanswered pre-sale questions, abandoned carts, and manual follow-up nobody has time for.

Stack

  • React Three Fiber
  • AI agent
  • n8n
  • E-commerce

The build

  • 3D product experience that makes the store worth remembering.
  • AI agent trained on the catalog — answers product questions, recommends, captures leads around the clock.
  • n8n workflows for order follow-ups, inventory alerts, and cart recovery.
  • Instrumented from day one: agent response time, chats handled, and conversion are being measured live — the numbers on this page will be real, or absent.

Status & outcome

This build runs exactly like the service offering: 3D craft on the front, agentic automation behind it. Real metrics will publish here as they are collected.

Your build could be story number four.

View Selected Work

Available for high-impact roles, technical leadership & contract architecture