Setu

Overview

Started as a site-level documentation digitisation tool for Military Engineering Services (MES) and evolved into a full construction project management platform covering planning, progress tracking, inventory management, billing, and team communication.

Construction Hierarchy

Every construction project breaks down into structures (pillars, spans, blocks) and further into activities (foundation pour, rebar installation, curing). This three-level hierarchy is the backbone of the platform — progress, inventory, scheduling, and reviews all operate at the activity level, while rollups give project-wide visibility.

RCPSP-Based Planning

Construction projects have resource constraints — a single crane can't be at two structures simultaneously. Activities have dependencies — you can't pour a deck before the pillars are up. The planner uses Resource-Constrained Project Scheduling (RCPSP) with Critical Path Method (CPM) to produce a feasible schedule that respects both constraints, and highlights the critical path — the sequence of activities where any delay delays the entire project.

Progress Analytics

Site engineers submit daily progress updates for each activity. These feed into analytics that compare planned vs actual completion, surface delays early, and track completion rates across the project. When an update shows a site falling behind, the variance is immediately visible — enabling corrective action before deadlines slip.

Inventory Tracking

Materials flow from central warehouse to site to activity. Every allocation and consumption is recorded and tied to a specific activity, so at any point you can see where materials are, how much has been used, and whether the numbers add up. Full reconciliation across all levels ensures nothing goes unaccounted.

Multi-Step Review Workflows

Every progress or inventory update goes through configurable approval stages. A site engineer submits, a junior engineer reviews, and a senior engineer gives final approval. At any stage, the reviewer can reject with a correction note — the submitter corrects and resubmits. This ensures data accuracy without bottlenecking the process, and every decision is recorded.

E2E Encrypted Communication

All client-server communication is end-to-end encrypted using Diffie-Hellman key exchange. Each client establishes a shared secret with the server, and every payload — messages, file attachments, progress updates, and all data — is encrypted before leaving the device. Nothing travels in plaintext. In-app messaging with file attachment support is built on top of this encrypted layer.

In-App Messaging

Project communication is built directly into the platform — no need to switch to WhatsApp or email. Team members can message within project context with full file attachment support for sharing site photos, documents, and reports. All messages flow through the E2EE layer, so conversations stay private and on-record.

AI Copilot

An AI copilot assists users across the platform. For data entry, it suggests values and auto-fills fields based on historical patterns, reducing manual effort and errors during daily updates. For project oversight, it provides summaries and insights across project data — surfacing trends, flagging anomalies, and helping decision-makers stay on top of progress without digging through raw numbers.

Audit Trail

Every action that updates any read model is tracked — submissions, approvals, rejections, corrections, and system-generated updates. The audit trail ensures full visibility into who did what and when, which is critical for government construction projects where accountability and documentation are non-negotiable.

Current Status

Completed a successful 3-month pilot and is now being adopted in the DGMAP division of Military Engineering Services (MES).

How It Was Built

The entire platform — backend, frontend, data models, E2EE layer, scheduling engine, and every demo on this page — was built using Claude Code. Not a single line of code was written manually. Architecture decisions, implementation, debugging, and iteration — all done through AI-assisted development.