Scroll down

Project Overview

CONNECTIS replaces spreadsheets and phone calls with one command center for fleet operators.

Outcomes

68 %

Faster triage, 7.9 → 2.4 min

31 %

Fewer days with vehicles off road

Challenge

CONNECTIS — a fleet operations manager in a glass-walled meeting room

Transform fragmented data into prioritized actions to prevent downtime.

Process

01AI-driven Product Strategy and Framing
Research deliverables: market research, user personas, journey mapping, problem framing

Accelerated market research, user personas, journey mapping, and problem framing using AI. Compressed discovery and scoping from months to days without sacrificing depth.

02Tokenized Design System and Architecture

Built sitemaps, user stories, and lo-fi wireframes into a high-fidelity Figma component library with Level-II token architecture for light and dark modes. Delivered a systematic design foundation structured for fast, accurate AI translation.

03Figma-to-Claude Integration and Documentation

Used Claude to parse Figma layouts, design tokens, and visual direction directly. Automated generation of PRDs, user flows, and precise UI reference specs from a single source of truth.

04AI-assisted Build and PoC Delivery

Fed design data straight into code-generation pipelines for instant iteration, then immediately usability-tested the output. Replaced slow handoffs to build and validate a working PoC in record time.

Deliverables

Hi-fi desktop screens from sign-in to the asset page, in light and dark modes, with 100% auto-layout structure.

Geometric precision: a Michroma wordmark, zero decoration, and one connected network mark for the fleet.

A 2-tiered token system — a foundational reference layer feeding a semantic layer built for the product.

Build

Full auth and role-based access, sign-in to sign-out. Supabase simulates a fleet of 50+ vehicles.

394

Commits, One Month

76 KB

Initial JS Bundle

I use Glacier Flow, a Claude Code template: folder routing, a knowledge base, and gated tasks. All my work goes through the CLI in three steps:

How one change moves through Glacier Flow: ask and plan, draft the task and approve it — revising until the answer is yes — then build and commit.
Ask & PlanTask & ApproveBuild & Commit
No? Revise
GitHub

What I've Learned

A production-ready PoC built, tested, and validated in record time.

Check on Behance