Chainsilience AI

Overview

Chainsilience AI is a supply-chain risk intelligence platform I developed alongside a team of four for IDSOL2026, aimed at helping SMEs, startups, and semiconductor companies detect, assess, and respond to external disruptions — geopolitical events, tariffs, export restrictions, natural disasters, shipping disruptions, material shortages — before they become costly operational problems.

Approach

The platform models supply chains with a knowledge graph and a digital twin, used for mapping supply-chain paths, disruption detection, impact simulation, risk scoring, and mitigation strategies.

For retrieval, I built a hybrid RAG pipeline using LangChain, combining FAISS vector search and sentence-transformer embeddings with BM25 keyword retrieval, so relevant signals aren't missed by either approach alone. LLM agents sit on top of that retrieval layer to judge news relevance, verify claims, reason about risk, and generate recommendations.

Design Philosophy: Auditable, Not a Black Box

The system is built around a human-in-the-loop workflow, not full automation. AI agents gather and analyze external information, connect disruptions to the specific supply-chain dependencies they affect, assess potential business impact, and generate recommended response actions and draft communications — all for human review and approval before anything happens.

A core focus of the project is making AI-generated decisions traceable: every recommendation is backed by its underlying evidence and structured into an actionable workflow, so a decision-maker can see why an alert matters and what to do about it, rather than trusting an opaque score.

Team & Traction

Built with a team of four; the platform was piloted with a real company during development.

Future Improvements

• Integrate approved workflows with enterprise communication systems (e.g. Gmail) for AI-assisted execution while keeping a human in the approval loop
• TODO: add source code link / live demo