Actus Logic
TL;DR: To solve the challenge of non-deterministic human workflows in high-stakes environments, I architected a distributed process governance engine using a Serverless RPC architecture and Supabase Edge Functions. By implementing a version-clocked state machine and offloading VLLM-based visual verification to the edge, I built a system capable of enforcing atomic state transitions and real-time process integrity on low-compute, high-latency mobile hardware.
Introduction
Actus Logic is an AI-powered execution engine designed for the deskless workforce. While ERPs handle planning, the physical act of work remains a "data black hole." Actus Logic fills this void by providing Operational Genealogy—an immutable, second-by-second record of how work was performed. It’s built specifically for high-compliance industries like construction and manufacturing where "faking safety" or "procedural drift" leads to catastrophic costs.
Design Approach
"Junk Phone" Optimization: Deskless work happens on low-end, cracked-screen Androids in crawlspaces. I optimized the UI for high-latency environments, using a "Link-to-Run" model that bypasses App Store friction and works instantly in any mobile browser.
Cognitive Load Engineering: Workers shouldn't look at screens; they should look at their work. I designed a voice-first interaction model with "glanceable" UI states so workers can receive instructions and verify steps hands-free.
The Verified Truth: Unlike passive checklists, Actus Logic requires Visual Verification. I integrated VLLM (Vision-Language Models) to analyze photos in real-time, ensuring a safety switch is actually "Off" before the next step unlocks.
Technical Stack
Infrastructure & Edge: Supabase Edge Functions (Deno/TypeScript) for globally distributed, low-latency execution. This allows the "Execution Layer" to live as close to the field worker as possible, minimizing round-trip times on poor cellular connections.
Communication Layer: PostgreSQL RPC (Remote Procedure Calls) to trigger complex database logic directly from the client. This bypasses traditional REST overhead, enabling high-speed state transitions and data integrity checks.
Intelligence: VLLMs & Custom AI Agents integrated at the Edge for real-time visual inspection and "Self-Patching" logic, where the system identifies bottlenecks and auto-updates SOPs.
Data Integrity: PostgreSQL (Supabase) utilizing a hardened, append-only ledger architecture to ensure "Operational Genealogy" remains immutable and audit-ready.
Frontend: React 18 & Tailwind CSS, featuring a custom state machine designed to handle complex branching logic and resilient data syncing on "low-end" hardware.
