About QuestFeed

Deterministic Intelligence for Complex Systems

QuestFeed builds AI-native decision support systems that model how risk cascades through complex, interconnected networks. We started in cybersecurity, expanded to regulatory compliance, and the architecture adapts to every graph-shaped problem that follows.

How We Got Here

It started in cybersecurity.

We noticed that vulnerabilities don't matter one at a time. They chain together into multi-step attacks. So we modelled those chains. 507 million threat intelligence records. 184 machine learning models. 422,000+ mapped attack chains across 32.6 million graph edges. Our forecast model predicts which new vulnerabilities will be weaponised at 0.837 AUC. The industry standard scores 0.500 on the same task.

That became AuditROI.

Then we pointed the same engine at regulation.

Regulatory obligations cascade just like vulnerabilities. A single compliance failure can trigger enforcement across dozens of dependent rules, and nobody models that. So we decompose regulatory text into atomic obligations, build dependency graphs, compute cascade probabilities from real enforcement data, and score every rule across five risk dimensions. Today: 145 frameworks, 115,000+ scored obligations, 61,000+ Bayesian conditional probability tables.

That became AuditDSS.

Along the way, we solved context retrieval.

Conventional RAG approaches weren't good enough. So we built a context engine that uses relational database structures to optimise how document chunks are sized, overlapped, and retrieved. 40% fewer tokens consumed. Significantly higher retrieval accuracy. That became ContextROI — powering our platforms under the hood and available as a standalone product.

Now we're building the reasoning layer on top.

ReasonROI is a deterministic agentic system. The graph reasons, the AI communicates. Hallucination isn't just reduced — it's architecturally impossible at the decision layer, because every reasoning step follows a scored graph, not a language model prediction. This enables self-verifying policy generation: the system writes a compliance policy, then scores it against the full obligation graph to check that nothing is missing.

The Core Engine

We built an engine that reads flat documents, unstructured data, and complex rule systems and turns them into scored, navigable intelligence networks.

How It Works

  1. 1 Takes regulations, threat data, or technical specifications and breaks them into their smallest atomic pieces
  2. 2 Every requirement, condition, prohibition, and threshold becomes a node in a relational database
  3. 3 Builds those nodes into Directed Acyclic Graphs with Bayesian reasoning layers on every edge
  4. 4 Every connection carries a learned probability, and risk cascades through chains the same way failures cascade through real systems

What Makes It Different

Not an LLM wrapper

No language model makes decisions. Production ML models, Bayesian networks, and graph algorithms do the reasoning. Deterministic and auditable.

Deterministic & auditable

Every score traces back to its source. Reproducible. No hallucinations, no black boxes.

Domain-agnostic architecture

Anything modelled as interconnected nodes with cascading probabilities is a target. New verticals adapt in weeks, not months.

Leadership

The Founder

Dr. Ali Ershadi. Systems thinker, generalist, and builder. Not a cybersecurity specialist or a regulatory expert — someone who sees patterns across domains and has the technical range to turn them into products.

PhD, Civil Engineering (UNSW)

Remote Sensing & Spatial Science — complex systems modelling

5 Years Postdoctoral Research (KAUST, Saudi Arabia)

Remote sensing and complex systems at one of the world's top research universities

Research Scientist (ANU)

Australian National University

Full-Stack Developer Since 2015

20+ years in data analytics & ML. Hands-on builder — not just strategy, but working systems.

"Every regulation, every network, every supply chain is a graph. Nodes connected by rules, dependencies, and probabilities. Risk doesn't sit in one node — it flows through chains. Detecting risky chains before they break is the real Return on Intelligence."
Dr. Ali Ershadi - Founder & CEO of QuestFeed

Dr. Ali Ershadi

Founder & CEO, QuestFeed

Our Platforms by the Numbers

Two production DSS platforms processing real-world cybersecurity and regulatory compliance data at scale.

AuditROI — Cybersecurity

507M+

Threat intel records

184

ML models

422K+

Attack chains mapped

266

Automated scanners

AuditDSS — Regulatory Compliance

145

Regulatory frameworks

115K+

Scored obligations

61K+

Bayesian CPTs

1,414

Enforcement actions

FAQ

About QuestFeed

What is QuestFeed and what does the company do?

QuestFeed is an Australian technology company that builds AI-powered decision support systems. We take complex, interconnected data — regulations, threat intelligence, technical specifications — and turn them into scored, navigable intelligence networks using graph-based reasoning with Bayesian probability layers.

Who founded QuestFeed and what is their background?

QuestFeed was founded by Dr. Ali Ershadi, who holds a PhD in Remote Sensing & Spatial Science. He has 20+ years of experience in data analytics, machine learning, and complex systems modelling, with postdoctoral research at KAUST (Saudi Arabia) and experience as a research scientist at ANU. He's a full-stack developer who sees patterns across domains and turns them into products.

What makes QuestFeed's AI technology different?

Our AI is built on production ML models, Bayesian networks, and graph algorithms — not language models. LLMs only handle document ingestion and communication. The graphs reason, the models score. That means no hallucinations at the decision layer, deterministic outputs, full auditability, and a fraction of the running cost of LLM-dependent systems.

How does the graph-based reasoning engine work?

The engine reads flat documents and complex rule systems, breaks them into atomic pieces, and builds them into Directed Acyclic Graphs with Bayesian reasoning layers on every edge. Each connection carries a learned probability, and risk cascades through chains the same way failures cascade through real systems. Every score traces back to its source.

What products has QuestFeed built?

Two platforms are live: AuditROI (auditroi.com) for cybersecurity threat intelligence with 266 scanners and 507M+ threat records, and AuditDSS (auditdss.com) for regulatory compliance intelligence across 145 frameworks with 115K+ scored obligations. InvestDSS and TenderDSS are in development.