At re:Invent 2025, AWS unveiled Nova Forge, a breakthrough platform that lets organizations build their own frontier-grade AI models—without needing a research lab, a GPU cluster, or a team of foundation-model scientists.

For the public sector, enterprises, and regulated industries, this could just be the most important AI announcement of the year.

Nova Forge enables something that organizations have struggled to achieve with traditional fine-tuning:

A domain-expert model that deeply understands your data—while preserving the reasoning power of a frontier model.

AWS calls these custom models “Novellas.” Unlike fine-tuned models that can overwrite or distort original capabilities, Novellas maintain the base model’s reasoning and instruction-following while gaining your organization’s proprietary context.

Book a 1:1 call to discuss your project

Why Nova Forge Matters

Most organizations fall into one of two buckets today:

1. They can’t afford to build custom models from scratch.

Training a large model requires millions of dollars in GPUs and months of specialized engineering. This has placed frontier-class models out of reach for all but the largest tech giants.

2. They rely heavily on fine-tuning off-the-shelf models.

Fine-tuning is useful, but it has limits:

• It can cause “catastrophic forgetting” of core model skills.

• It often produces brittle models that fail outside narrow prompts.

• It doesn’t allow deep domain integration at the foundational level.

Nova Forge bridges this gap.

You start with a Nova model checkpoint, then progressively blend in your own data—contracts, transcripts, case files, call logs, sensor data, or even simulation environments—alongside Amazon-curated datasets.

The result is a model with frontier intelligence + your institutional knowledge + your governance boundary.

Public Sector possibilities

Public agencies are the perfect fit for Nova Forge.

High-value data, trapped in documents

State procurement teams, justice departments, transportation agencies, and health services have decades of:

• RFPs

• contracts

• case documents

• regulatory filings

• FOIA requests

• audio/video transcripts

• inspections and field reports

Nova Forge can convert this history into decision-ready, domain-specific intelligence—while keeping all data inside your AWS boundary, meeting FedRAMP, HIPAA, CJIS, IRS-1075, or state-specific compliance requirements.

Improved decision-making

Imagine Novellas built specifically for:

• contract scope generation

• bid scoring

• case triage

• benefits eligibility

• policy analysis

• code enforcement review

• public records classification and redaction

These are not generic LLMs—they’re models that understand your rules, your terminology, your workloads.

Commercial Enterprise possibilities

Across industries—insurance, manufacturing, healthcare, energy, and finance—Nova Forge unlocks capabilities previously available only to hyperscalers.

Insurance

A Novella trained on:

• claims histories

• medical records

• adjuster notes

• fraud indicators

…becomes a claims-domain expert that never leaves your VPC.

Manufacturing & Industrial Operations

Blend:

• sensor streams

• equipment manuals

• maintenance logs

• incident reports

…to produce predictive maintenance models with domain-aware reasoning.

Retail & Customer Experience

Train on:

• call center transcripts

• product data

• service interactions

• customer journeys

…to create CX copilots that reflect your brand and knowledge base.

Reinforcement Learning Gyms: A Hidden Superpower

Nova Forge also includes reinforcement learning gyms, enabling organizations to train agents inside simulated environments:

• policy simulations

• emergency response scenarios

• procurement workflows

• supply chain disruptions

• moderation or safety pipelines

Reddit is already using Nova Forge for community moderation by injecting subreddit-specific knowledge into a custom AI model.

This is an extraordinary capability for:

• emergency management

• transportation

• utilities

• logistics and operations

• public safety

Why This Matters Now

Generative AI adoption has been bottlenecked by two issues:

1. Trust – “Will my data leak or retrain a public model?”

2. Performance – “Can a general-purpose model really understand our domain?”

Nova Forge answers both:

✔ Proprietary data never leaves your AWS environment
✔ Training happens with Amazon-curated datasets
✔ Foundational reasoning is preserved
✔ You get a model that reflects your workflows, rules, and context

This finally gives organizations the ability to move from pilots and demos to production-grade, domain-expert AI systems.

Where Metal Toad Can Help

As an AWS Advanced Consulting Partner with Generative AI, Machine Learning, and Public Sector competencies, Metal Toad helps organizations:

• Evaluate where Nova Forge fits into your roadmap

• Prepare and structure your proprietary data

• Build your first Novella model

• Integrate custom models into Bedrock applications

• Design reinforcement learning environments

• Ensure compliance and governance

• Deploy models into production systems (web, mobile, backend services)

If you’re exploring Nova Forge and want to understand its impact on your organization’s AI strategy, we’d love to talk.

👉 Contact us for a Nova Forge readiness consultation.