Prodigy AI Solutions

Prodigy AI Solutions has officially launched Verbis Graph SaaS on the Google Cloud Marketplace! 🚀

Verbis Graph acts as a reusable enterprise semantic context layer, translating raw unstructured documents into structured meaning so your AI agents can perform complex, multi-hop reasoning with absolute trust.

Why leading enterprise teams are deploying Verbis Graph today:
🛡️ Unbroken Evidence Paths: Significantly reduce hallucination risk and terminology ambiguity. Verbis Graph provides full answer traceability, linking every response directly back to the original source document passages.
📊 Supercomputer-Validated Quality: As an NVIDIA Inception partner, we've rigorously benchmarked our GraphRAG engine on the legendary CINECA Leonardo HPC supercomputer infrastructure in Italy—achieving up to a 92.26% retrieval hit rate and 95.3% financial coverage.
⚡ Frictionless Procurement: Skip the lengthy vendor-verification and legal onboarding queues. Procure instantly using your pre-committed GCP spend (CUDs) and consolidate your billing onto a single, existing Google Cloud invoice.

Stop flattening your enterprise data. Give your AI stacks the structural relationships they deserve!

🩺 We’re pleased to share that Prodigy AI Solutions has joined the EIT Health Innovators Community and, by extension, the wider EIT Health network.

Being part of a community connecting researchers, academics, healthcare professionals, scientists, entrepreneurs, investors and other health innovators is particularly valuable as we enter the next stage of CohortForge.

We’re preparing to move from development toward testing and evaluating CohortForge with real-world datasets together with clinical partners. This next phase will help us understand how the platform performs in real clinical research scenarios and where we can continue improving it.

🔬 Learn more: med.prodigy-ai-solution com

🚀 Another big milestone for Verbis Graph: approved for publication in the NVIDIA App Catalog.

We’re excited to share that Verbis Graph has been approved for publication in the NVIDIA App Catalog, in the Enterprise Applications section. The NVIDIA App Catalog brings together applications, tools, and services built on NVIDIA platforms for areas including AI, cloud computing, data science, and enterprise workloads.

For us, this milestone is important because Verbis Graph is being built specifically for the challenges enterprises face when bringing AI into real-world environments: connecting fragmented knowledge, improving retrieval, grounding AI responses in evidence, and making enterprise information more explainable and useful for AI agents.

🕸️ Verbis Graph transforms documents and enterprise data into connected knowledge that AI systems can retrieve and reason over. Being approved for publication in the NVIDIA App Catalog gives us another channel to bring that technology closer to organizations building on NVIDIA's ecosystem. It also comes at an exciting point in our journey.

Prodigy AI Solutions is a member of NVIDIA Inception, and we’re happy to continue growing within the NVIDIA ecosystem as we develop Verbis Graph for increasingly demanding enterprise AI workloads. This milestone joins several important steps we've taken recently - from benchmarking graph-enhanced retrieval on HPC infrastructure to expanding Verbis Graph's availability across enterprise cloud marketplaces. There is still plenty to build, test, and learn, but today we're taking a moment to celebrate this one.

🚀 The complete Verbis Graph evaluation report is now available.

We’ve been evaluating our graph-enhanced retrieval technology on HPC infrastructure using standardized datasets including NFCorpus and SciFact. The latest results include an improvement in NFCorpus MRR from 0.303 to 0.578, while our first SciFact evaluation achieved a 98% Hit Rate and 98% Mean Recall.

These experiments help us understand how GraphRAG and knowledge graph retrieval can find and rank relevant evidence for more accurate, grounded, and explainable enterprise AI.

📄 Read the full evaluation: https://verbisgraph.com/uploads/VerbisGraph-Evaluation%20of%20Graph-Enhanced%20Retrieval%20on%20HPC%20Infrastructure.pdf?utm_source=facebook&utm_medium=social&utm_campaign=evaluation_report_15092026

🎉 We’re happy to share that Prodigy AI Solutions has been accepted into the MongoDB for Startups program!

It’s another meaningful step for our team as we continue developing Verbis Graph (verbisgraph.com) and building scalable AI solutions for real-world enterprise use cases. 🚀

When enterprises evaluate an AI solution, accuracy is only part of the conversation. Security, reliability, architecture, operations and deployment risk matter just as much.
That's why we're pleased to share that Verbis Graph Solution (verbisgraph.com) has successfully completed the AWS Foundational Technical Review (FTR). What does that actually mean for a customer considering Verbis Graph on AWS?
🔐 Greater security confidence. Security architecture and applicable AWS best practices are part of the technical review, helping customers evaluate the solution with more confidence.
⚙️ Production readiness. Enterprise AI needs to keep working beyond a proof of concept. Reliability, operational practices and architecture matter when AI becomes part of real business processes.
☁️ AWS technical validation. FTR involves an AWS technical review rather than simply being a badge we assign to ourselves. For organizations with structured technology and vendor-review processes, that distinction matters.
📉 Lower adoption friction. Having completed AWS's technical review can give architecture, security and procurement teams additional information when assessing the solution and its deployment on AWS.
For us at Prodigy AI Solutions, the milestone also reflects something we strongly believe: building enterprise AI isn't only about making models smarter. The surrounding architecture has to be secure, observable, reliable and ready for real-world workloads.
We’re continuing to build Verbis Graph with that principle in mind, and passing the AWS FTR is an important step on that journey. 🚀

This article gave us something to think about.

As AI agents become more autonomous, we spend a lot of time talking about what they can do: use tools, access systems, coordinate with other agents, execute tasks, and make decisions with less human involvement.

But there is another side to that progress.

What happens when an autonomous agent can also replicate itself—or when the infrastructure around it wasn't designed with that possibility in mind?

The article below discusses the risks of autonomous replication by AI agents and, in particular, security gaps that may exist in newer cloud environments.

What made this interesting for us is that it connects closely with our own experience at Prodigy AI Solutions.

As we build our orchestration agent and specialized sub-agents, we're increasingly finding that some of the most important engineering decisions aren't about making agents more capable.

They're about deciding where autonomy should stop.

What can an agent access?

What can it create?

Which actions require explicit permission?

Can it create or invoke another agent?

What credentials can it use?

How do we prevent one compromised agent from becoming a problem for the rest of the system?

And perhaps most importantly: how do we make all of this observable and controllable by a human?

These are subtle questions. There isn't always a simple technical answer, and we're learning from them as we build.

Our current thinking is that autonomy should never automatically mean unlimited authority. The more capable an agent becomes, the more carefully its permissions, boundaries, identity, and infrastructure need to be designed.

The article is worth reading if you're building or deploying autonomous agents:

https://nai500.com/blog/2026/09/risks-of-autonomous-replication-by-ai-agents-raise-concerns-security-gaps-of-new-cloud-service-providers-need-urgent-attention/

We'd also genuinely like to hear from others building in this space.

Where do you draw the line between useful agent autonomy and too much autonomy?

Improving AI-agent accuracy should not require an unlimited infrastructure budget.

Enterprise teams often try to make agents more reliable by using larger models, longer context windows, additional reasoning steps and repeated retrieval calls.

This increases infrastructure and token costs—but does not necessarily solve the underlying problem.

If an agent receives incomplete, disconnected or irrelevant evidence, more compute may simply produce a more elaborate incorrect answer.

Verbis Graph addresses the problem at the retrieval layer.

It transforms private enterprise documents into a connected, citation-backed knowledge graph and combines:

• semantic retrieval for meaning
• lexical retrieval for exact terminology
• graph traversal for entities and relationships
• multi-hop retrieval across multiple documents
• ontology support for enterprise-specific definitions
• citations linking answers to original evidence

Instead of filling the model’s context window with many loosely related passages, Verbis Graph helps retrieve a more focused set of connected and verifiable evidence.

This can help enterprises reduce AI infrastructure costs by:

• limiting unnecessary prompt tokens
• reducing repeated retrieval and generation attempts
• reusing one indexed knowledge layer across multiple agents
• supporting provider-agnostic and locally deployed models
• reserving larger models for tasks that genuinely require them

At the same time, agents receive the relationships, terminology and source evidence needed to produce more grounded and accurate answers.

During our CINECA Leonardo evaluations with a locally served 32B model, Verbis Graph achieved a 98% retrieval hit rate and 0.721 MRR on SciFact, together with 95.3% retrieval coverage and 0.584 MRR on FiQA.

These are retrieval results rather than a completed end-to-end cost benchmark, but they support an important direction: stronger enterprise AI may come not only from larger models, but from giving appropriately sized models better evidence.

Verbis Graph does not make the model bigger.

It makes the context more relevant, connected and verifiable.

https://verbisgraph.com/

AI for science needs reasoning—not just data. 🔬

AlphaFold demonstrated what AI can achieve with decades of carefully curated scientific knowledge. But most fields do not have comparable datasets. Their evidence is scattered across papers, experiments and institutional systems—and results may be incomplete, inconsistent or difficult to reproduce.

The next breakthrough could come from scientific AI agents that connect evidence, use specialised tools, evaluate uncertainty and continuously refine their conclusions.

Verbis Graph could support these agents as a traceable knowledge layer, connecting relationships across scientific sources and retrieving grounded evidence with citations.

At Prodigy AI Solutions, we are also completing a TRL 5 generative-AI model that transforms scarce, incomplete and imbalanced medical-imaging data into privacy-conscious synthetic cohorts. It is designed to support more representative AI development and help prepare models for rigorous local clinical validation.

We are open to collaborations with universities, research centres and healthcare or life-science organisations.

⚠️ What if one false relationship could influence multiple AI answers?

GraphRAG helps AI connect entities and relationships across documents. But those connections can also become an attack surface.

**GRAGPoison** is a research attack that introduces plausible false information into source documents. During indexing, a GraphRAG system may extract those deceptive relationships and incorporate them into its knowledge graph.

The attacker can then reinforce one poisoned relationship with supporting content, increasing the possibility that it will appear across multiple retrieval paths.

In selected experiments, researchers observed attack-success rates as high as **98%** while using up to **68% less poisoning text** than the comparison attack.

This does not mean every GraphRAG system has a 98% vulnerability rate. It does mean that enterprises need to protect the integrity of the entire knowledge lifecycle.

A safer GraphRAG architecture needs:

🔹 Controlled document ingestion
🔹 Trusted and traceable sources
🔹 Restricted modification rights
🔹 Monitoring of graph changes
🔹 Quarantine and rollback procedures
🔹 Human review for sensitive decisions

Verbis Graph (verbisgraph.com) supports controlled and traceable retrieval from approved enterprise documents, with authorised workspaces, citations, graph visualisation, optional ontology constraints and customer-controlled deployment.

It provides an inspectable foundation—but no responsible provider should claim that graph poisoning is completely solved.

The next question for enterprise AI is not only: “Is this information relevant?”

It is also: **“Where did it come from, and which AI decisions now depend on it?”**