Two Products.
One Compliance Journey.
Built for Nuclear-Grade Operations.
Start with Magic Lens — the AI research co-pilot that accelerates NRC compliance documentation to minutes. Built for nuclear decision traceability, compliance acceleration, and operational anomaly visibility. Scale to PulseFINE™ — the full federated AI infrastructure for anomaly detection, dual simulation, and sovereign intelligence at the edge.
Technology Development Supported By
Start Small. Scale Sovereign.
Magic Lens is the entry wedge. PulseFINE™ is the full platform. Both are designed to work together — one accelerates compliance documentation today, the other transforms how your facility thinks, predicts, and operates for decades.
Magic Lens
Browser-based AI research co-pilot. Surfaces regulatory citations, cross-references NUREG documents, and assembles NRC compliance packages in seconds — not days.
PulseFINE™
Edge-resident federated AI. Real-time anomaly detection, dual simulation workflows, cross-site forecasting — no data leaves the facility. Built for SMRs and critical infrastructure.
How Magic Lens and PulseFINE™ Differ
Two products. Different problems. Complementary capabilities. Choose your entry point based on where your facility is today — and where it needs to go.
| Feature / Dimension | 🔍 Magic Lens | ☢️ PulseFINE™ |
|---|---|---|
| Deployment & Architecture | White Paper Coming Soon! | |
| Deployment model | Chrome browser extension | Edge-resident bare-metal nodes (Supermicro + NVIDIA GPU) |
| Infrastructure required | ✓ None — browser only | On-prem or colocation GPU cluster |
| Data sovereignty | ⚠ Federated research, no raw ops data | 🔒 Full — zero data exfiltration |
| Air-gap compatible | — Requires browser internet | ✓ Fully air-gap capable |
| Cloud dependency | ⚠ Federated web layer | ✓ No cloud dependency |
| Core Capabilities | ||
| NRC compliance documentation | ✓ Primary use case | ⚠ Via PulseTRACE audit layer |
| Real-time anomaly detection | — Not applicable | ✓ Core capability |
| Dual simulation workflows | — Not applicable | ✓ Predictive + operational simulation |
| Cross-site federated forecasting | — Not applicable | ✓ Multi-site without raw data sharing |
| NUREG / ADAMS document search | ✓ History + Discourse lenses | ⚠ Via compliance layer |
| Predictive maintenance | — Not applicable | ✓ Edge AI inference |
| Expert network routing | ✓ Federated Research Network SMEs | ⚠ Internal expert layer |
| Compliance & Audit | ||
| 10 CFR Part 50 alignment | ✓ Built-in citation engine | ✓ PulseTRACE audit layer |
| Immutable audit logs | ⚠ Session provenance chain | ✓ Full model lineage + decision rationale |
| Explainable AI (XAI) | ⚠ Source citations per output | ✓ Full explainability layer |
| LER / CAR package assembly | ✓ Automated in <15 seconds | ⚠ Via compliance integration |
| Investment & Timeline | ||
| Investment range | $50K – $200K / year | $5M – $15M build |
| Time to first value | Days to weeks | Months (phased pilot deployment) |
| Procurement complexity | ✓ Low — SaaS subscription | High — infrastructure procurement |
| Ideal buyer | Compliance officer, CNO, NRC liaison | CTO, VP Engineering, DOE/national lab |
| Ideal stage | Immediate compliance acceleration | Full operational AI transformation |
Two Products. One Journey.
Magic Lens and PulseFINE™ are designed to work in sequence. Start with compliance acceleration. Scale to operational sovereignty.
🔍 Magic Lens
A federated AI research co-pilot built for nuclear compliance engineers. Surfaces regulatory citations, cross-references NUREG documents, and assembles LER submission packages in seconds — directly in your browser.
- ✓Four research lenses: History (regulatory), Discourse (NRC ADAMS), Multimedia (national labs), Expert (credentialed SMEs)
- ✓Automated ECCS compliance package assembly in under 15 seconds
- ✓10 CFR Part 50 citation engine with NUREG-0800 and NUREG-1022 cross-reference
- ✓Project Memory: links related findings across a compliance session
- ✓Full provenance chain on every output — document ID, version, retrieval timestamp
- ✓Federated Research Network routing to ORNL, INL, Sandia, PNNL readouts
- ✓Chrome extension — no infrastructure, no IT procurement, live in days
☢️ PulseFINE™
Federated Intelligence for Nuclear Engineering. Purpose-built for SMRs — real-time anomaly detection, dual simulation workflows, and cross-site forecasting without centralizing sensitive data or relying on public cloud AI.
- ✓Edge-resident AI agents monitoring reactor and infrastructure telemetry in real time
- ✓Federated forecasting across SMR sites without sharing raw operational data
- ✓Secure anomaly detection inside colocation or on-prem environments
- ✓GPU-optimized inference (Supermicro bare-metal + NVIDIA) tuned to operational workloads
- ✓PulseTRACE: model lineage, decision rationale, immutable audit logs for regulators
- ✓Dual simulation workflows: predictive maintenance + early-warning signal generation
- ✓No centralized cloud dependency — no data exfiltration risk — air-gap capable
How Each Product Handles an ECCS Failure
A real-world scenario: the Containment Isolation Valve Train A records 6.22 sec actuation vs. 5.0 sec Tech Spec limit. Here is how each product responds.
Identify the Anomaly
Lensy highlights the out-of-spec ECCS value in red inline. History lens instantly surfaces the applicable 10 CFR clause and NUREG-0800 standard. Discourse lens cross-references NRC ADAMS for precedent LERs with similar valve timing failures.
~3 secondsFind Related Surveillance Data
Project Memory connects Train A failure to previously highlighted Train B data in the same session. Multimedia lens queries ORNL, INL, and Sandia for approved readouts from comparable reactor configurations.
~4 secondsPull the Regulatory Clause
History lens surfaces 10 CFR 50 Appendix B and NUREG-0800 Section 6.2 acceptance criteria. Expert lens routes to a credentialed SME for CAR threshold confirmation. Discourse lens cross-references NUREG-1022 Rev. 3 for event reporting guidance.
~4 secondsBuild the Submission Package
Lensy's "Collect into a Skill/report" action aggregates all highlighted findings, tool outputs, and source citations into a single structured package with full provenance chain — document ID, version, retrieval timestamp.
<15 seconds totalPre-Anomaly Detection
Edge AI agents monitoring valve actuation telemetry in real time detect the developing timing drift before it reaches the 5.0 sec threshold. Early-warning signal generated at 4.7 sec — before the out-of-spec event occurs.
Continuous monitoringFederated Cross-Site Analysis
Federated forecasting queries comparable valve performance data across all federated SMR sites — without transmitting raw operational data. Identifies whether this is a site-specific or fleet-wide pattern.
Real-time inferenceDual Simulation Workflow
Runs predictive simulation: what is the probability of Train B failure given Train A drift pattern? Generates early-warning signal with confidence interval. Operational simulation models corrective action scenarios.
GPU-accelerated inferencePulseTRACE Audit Package
Immutable audit log captures model lineage, decision rationale, and all inference steps. Compliance-ready output for NRC oversight. Every decision is explainable, traceable, and auditable — no black-box outputs.
Continuous audit trailHow PulseFINE™ Is Built
Five layers of sovereign, auditable, edge-resident AI — each designed to meet nuclear-grade operational and regulatory requirements without centralizing sensitive data.
Edge Nodes
Supermicro bare-metal + NVIDIA GPUs. Local inference, short-horizon prediction, secure data vaults. No data leaves the facility boundary.
Control Plane
PulseDNA federated learning coordination, policy-as-code enforcement, secure aggregation. Models learn collectively; raw data stays local.
Operational AI Layer
Predictive maintenance, anomaly detection, early-warning signals, dual simulation workflows. GPU-optimized for operational workloads.
PulseTRACE
Model lineage, decision rationale, immutable audit logs. Explainable AI for regulators and oversight bodies. Every inference is traceable.
Air-Gapped Execution
Full air-gap capability. Zero cloud dependency. Zero data exfiltration risk. Designed for national security, safety, and energy regulations.
Who Each Product Serves
Magic Lens and PulseFINE™ serve different buyers at different stages of the nuclear AI adoption curve. Both are designed for the same end goal: operational sovereignty.
NRC Compliance Officers
Accelerate LER and CAR package assembly. Surface applicable regulatory citations in seconds. Build audit-ready provenance chains for every submission.
Magic LensSMR Operators
Real-time anomaly detection across reactor and infrastructure telemetry. Federated forecasting across distributed sites without centralizing sensitive operational data.
PulseFINE™Chief Nuclear Officers
Reduce regulatory liability with AI-assisted compliance documentation. Demonstrate to the NRC that every decision is grounded in verified, cited knowledge.
Magic LensNational Labs & Energy Agencies
Sovereign AI infrastructure for ORNL, INL, Sandia, PNNL. Edge-resident intelligence that meets DOE security requirements. No public cloud exposure.
PulseFINE™Nuclear-Adjacent Infrastructure
Critical infrastructure operators who need sovereign AI without nuclear-specific licensing. Secure industrial facilities, energy agencies, and defense-adjacent operations.
PulseFINE™Compliance Research Teams
Engineers who spend hours manually searching NRC ADAMS, NUREG documents, and national lab readouts. Magic Lens reduces research time from hours to seconds.
Magic Lens"We are not building a search engine. We are building a co-processor for human reasoning — one that ensures every decision made at a nuclear facility is grounded in the best available verified knowledge. Magic Lens gets you there in weeks. PulseFINE™ gets you there forever."
Stephanie Joyce — Founder, PulseDNA
