Solyra
In chaotic environments, multiple actors report the same incident while others go unnoticed. Solyra is a decision-support augmentation layer that uses advanced probabilistic algorithms to group overlapping reports into single, uncertainty-aware “Event Nodes.” It replaces static, noisy maps with a dynamic “Living Map,” allowing coordinators to triage actionable incidents without the epistemic risk of manual deduplication.
From a static map to a Living Map.
In high-density refugee settlements, deterministic mapping fails: the same incident is reported many times over, while limited humanitarian resources are misdirected by noise. Manual deduplication carries real epistemic risk — the risk of erasing a genuine second event by assuming it's a duplicate.
Solyra makes the shift from deterministic mapping to an uncertainty-aware framework. It groups overlapping reports probabilistically, keeps human operators in the loop, and — crucially — exposes its own uncertainty rather than hiding it.
Red. High-confidence, high-severity Event Nodes demanding immediate, actionable response.
Yellow. Uncertain or emerging clusters — watch, verify, and let the bounds narrow before committing resources.
Green. Resolved or low-priority nodes, kept visible so nothing genuinely new is mistaken for a duplicate.
Probabilistic deduplication, engineered to scale.
Multi-Signal Similarity
A similarity model weighs multiple signals — space, time, and semantics — to judge whether two reports describe one event or two.
Graph-Based Clustering
Reports are clustered into Event Nodes on a graph, so overlapping observations collapse into a single, uncertainty-aware incident.
Approximate Nearest-Neighbor
ANN techniques overcome the computational bottleneck, keeping the Living Map responsive even at settlement scale.
Low-Bandwidth & Offline
Engineered for high-density, low-connectivity environments — operate offline or sync asynchronously without losing the thread.
The research behind Solyra.
Four documents by Dr. Sohaib Khan - from probabilistic theory to the developer blueprint and ethical framework.
Probabilistic Geospatial Deduplication for Decision Support
The foundational research: how uncertainty-aware deduplication transforms decision support in high-density refugee settlements, and why the shift from deterministic mapping prevents the misuse of limited humanitarian resources.
Operating the Living Map
For field coordinators and crisis responders: how to navigate the dynamic Living Map, interpret uncertainty bounds rather than static points, and manage Red / Yellow / Green triage — online or offline.
Inside the Deduplication Engine
The exact mathematical models and system layers behind Solyra — the multi-signal similarity model, graph-based clustering, and approximate nearest-neighbor techniques that translate complex data into actionable intelligence.
Managing Epistemic Risk at Scale
How Solyra balances rapid response with strict data protection — privacy-by-design with no PII required for core functionality, humans kept in the loop, and compliance with UN OCHA and GDPR guidelines.
“Even when the grid fails, our humanity — and our ability to help one another — should not.”
Most tech founders come from Silicon Valley. My perspective was forged in clinics, refugee contexts, and two decades of academic research into global health systems. As a Medical Doctor with a PhD in Public Health, I have spent my career studying how fragile systems break — and how to make them resilient.
