🦅 Executive Brief  ·  3DAgility

The AI that speaks
engineer.

Peregrine is the AI design assistant that understands what engineers are trying to build — in plain English — and recommends exactly the right components with plain English reasoning. Not spec search. Intent understanding.

📐 $90B+ Connector Market ✓ POC Audited & Validated 🧠 Translation Layer = The Moat 🔒 3DAgility Confidential
You don't ask Jarvis for the schematics. You tell Jarvis what you're building — and Jarvis figures out the rest.
— The Peregrine Vision
01 — The Problem

Engineers waste hours finding parts they could find in minutes.

Every electronic product starts with a connector problem. An engineer knows exactly what the application needs. But translating that into a spec, filtering 847 results, reading datasheets, and waiting two days for a FAE reply — that costs hours per decision. Multiplied across every connector on every board, it costs programs weeks.

The tools haven't changed in twenty years. Distributor search is keyword search. Manufacturer catalogs are PDFs. Cross-reference tools are lookup tables. None of them understand what the engineer is actually trying to build.

$90B+
Global connector market (2024)
$130B
Projected market by 2030
$500B+
Total electronic components market
Hours
Lost per design decision today
02 — The Product

Intent-first AI. Not spec search.

Peregrine's core breakthrough is the Translation Layer — a proprietary engine that maps what an engineer is building directly to the connector specifications required, without the engineer needing to know what those specs are.

See it in action

Live Conversation — Peregrine Design Assistant
E
Engineer
I'm building a charging module for an industrial IoT gateway. It's going on a DIN rail in a factory. Battery pack to main PCB. Dusty environment, vibration from machinery, and the maintenance crew needs to disconnect and reconnect it without tools.
P
Peregrine
I'm working from: industrial → IP-rated, dust protection · vibration → positive locking / high-retention · tool-free maintenance → push-pull or lever-actuated · DIN rail → standard mounting constraints

Before I give you parts — what's your voltage and current range?
E
Engineer
12V, 10A, through-hole PCB mount.
P
Peregrine
Top Pick: TE Connectivity AMP+ HVA 280 Series
Lever-actuated latching — maintenance staff disconnects without tools, no accidental pulls. 12A continuous, IP2X dust protection, designed for exactly this application. DigiKey: $2.14/unit · 3-week lead time · 847 in stock

Also Consider: Phoenix Contact COMBICON MC Series
Classic DIN rail choice. Tool-free spring cage termination, 16A rated, vibration-tested to IEC 60068. Slightly larger. $3.40/unit · 2,200 units in your ERP right now.

No dropdown menus. No datasheets until asked. The engineer described a problem — Peregrine turned it into specs, specs into parts, and parts into a recommendation with live data.

Three Personas. One Intelligence Engine.

Persona 01

End User — Design Assistant

Engineer · Procurement

Natural language → right part, every time. Cross-manufacturer recommendations with plain English reasoning. Live availability and pricing from connected distributor APIs. No spec knowledge required.

Persona 02

Distributor — Sales Accelerator

Mouser · Arrow · Regional Specialists

AI engine for customer interactions. Intelligent cross-reference. Proactive alternatives when parts are allocated. Customer-facing chat powered by the distributor's own inventory and preferred stock.

Persona 03

OEM — Product Intelligence

TE Connectivity · Molex · Amphenol · Hirose

AI layer on the OEM's own catalog. Application-driven search within their product line. Competitive positioning. Surfaces products customers didn't know existed — in the OEM's own portfolio.

How the Intelligence Works

USER INPUT "I'm building a wearable heart monitor. PCB to flex. Worn on wrist. FDA Class II." │ ▼ INTENT AGENT [GPT-4o structured output] Extract: application type · environment · electrical · compliance · missing info │ ▼ TRANSLATION LAYER ◀── 3DAgility's IP. This is the moat. Map: "wrist-worn" → <0.5mm pitch · "sweaty" → moisture protection · "FDA Class II" → UL94 + RoHS Knows the spec from the application — not from what the engineer types │ ▼ RETRIEVAL AGENT [Pinecone + Live APIs] Find: real parts from real catalogs with live pricing, stock, lead times, ERP data │ ▼ RESPONSE AGENT [GPT-4o streaming] Deliver: plain English recommendation with reasoning — streamed progressively
03 — Current State

POC built. Audited. Ready to rebuild right.

The Winjit POC validated the core approach — RAG architecture, intent understanding, real-time distributor data. It works in the happy path. A full technical audit by MAAX (June 2026) confirmed the concept is sound and identified exactly what needs to change to go from proof-of-concept to production.

★★★☆☆☆☆☆☆☆ Audit Rating: 3/10 for production readiness — but 10/10 for concept validation

✅ What the POC Proved

  • RAG architecture works for connector selection
  • Natural language → spec intent is mappable
  • Pinecone + OpenAI pipeline is solid
  • Real-time DigiKey / distributor data integration viable
  • Wiremasters (3M + TE catalogs) adds a second product line
  • Users can hold coherent multi-turn conversations with the AI

⚠️ What Gets Rebuilt

  • !Hardcoded API keys throughout codebase — critical security issue
  • !No test coverage (README templates literally say "TODO")
  • !Multiple conflicting version files — uncontrolled changes
  • !Raw OpenAI SDK calls instead of proper agent orchestration
  • !Prompts hardcoded in application code — not versionable
  • !No streaming UI — wall of text on response
The honest summary: this is a working prototype that proves the concept. It was never built to enterprise standards — but that's not the point. The point is that the idea works. Now we build it right.
🤖

LangGraph Orchestration

Replace raw SDK calls with a proper multi-agent state machine. Each agent has a specific job. Observable, debuggable, improvable independently.

Streaming Responses

Results stream progressively — no waiting. Feels fast even when processing is deep. Engineers see Peregrine thinking in real time.

🔌

Plugin Data Architecture

DigiKey, Mouser, Arrow, customer ERP — each is a plugin. New data sources added without touching core infrastructure.

🔒

Enterprise Security First

Azure Key Vault for all secrets. Entra ID for auth. Zero credentials in code. CORS locked down. Audit logging from day one.

04 — Competitive Advantage

The moat nobody else can cross.

Every major distributor will add an AI chatbot. The big OEMs are running LLM pilots. DigiKey has an AI team. None of them have built — and none of them can easily build — what Peregrine has.

The Translation Layer is the proprietary engine that maps application intent to connector specifications. It encodes decades of connector engineering expertise — the kind that lives in the heads of experienced FAEs. It knows that "wrist-worn" means <0.5mm pitch. It knows that "factory vibration" implies IEC 60068 retention requirements. That knowledge is not in any public dataset. It's in Peregrine.
🧠
Domain Intelligence

We understand the application, not just the catalog. Nobody else has this Translation Layer.

AI Execution Speed

We build and iterate faster than any traditional electronics company can move.

🔗
Live Data Integration

Real-time pricing, inventory, ERP data. Not a stale catalog that's wrong before it loads.

📈
Gets Smarter Over Time

Every engagement trains the Translation Layer. The moat deepens with every customer interaction.

💰
Revenue from Day One

Consulting-led model. Profitable immediately. Every engagement funds the next capability.

🔒
High Switching Costs

Once integrated into a customer's ERP and workflows, the relationship deepens every day.

✅ What We Have

Domain expertise — Mark, Bob, Joe — 15+ years TE Connectivity
Working POC (Winjit) — architecture and concept validated
Pinecone + Azure + OpenAI infrastructure proven
AI orchestration capability (MAAX)
Consulting model → immediate cash flow

❌ What Competitors Lack

Intent-to-spec Translation Layer
Deep connector domain knowledge encoded in AI
Multi-persona architecture — one engine, all three markets
Speed — traditional electronics companies move slowly
AI-native — they're bolting chatbots onto legacy keyword search
05 — Market Opportunity

A $500B market with 20-year-old tools.

The connector market alone is $90B annually. The broader electronic components market is $500B+. Every company in this space — OEMs, distributors, engineering firms, procurement teams — has the same intent-to-spec problem. None have solved it. The window is open. And it won't stay open.

🏭

Tier 1 OEMs

TE Connectivity, Amphenol, Molex, Hirose, JST. License Peregrine for their own catalog. Internal engineers use it; they offer it to customers as a value-added service.

High ACV Strategic
📦

Distributors

DigiKey, Mouser, Arrow, Avnet, regional specialists. White-label Peregrine as their AI search and recommendation engine for customer-facing interactions.

Volume Fast Path
⚙️

Engineering Firms / CMs

Design assistant for engineering teams. Cut component selection time from hours to minutes per decision. High-frequency usage drives high retention.

High Usage SaaS
🛒

Enterprise Procurement

Validate specs, find alternates, cut sourcing time. Connect directly to their ERP, approved vendor lists, and contracted pricing — live, not static.

ERP Integration Retention

The window is open — but it won't stay open. The distributors know this. DigiKey has an AI team. Mouser is experimenting. The big OEMs are running LLM pilots. None have cracked the core problem — they're all building better keyword search wrapped in a chat interface. We have the domain knowledge, the AI capability, and the speed advantage to get there first. That window closes in 18–24 months.

06 — Roadmap

Phase by phase to platform.

We start narrow and go deep. The End User Design Assistant for connectors is the beachhead — highest pain, clearest value, hardest to replicate. Once the Translation Layer is proven, every other category and persona is an expansion, not a rebuild.

Now → 6mo
Phase 1

🎯 End User Design Assistant — Connectors

Build the core product. End User persona. Connectors only. Intent Agent + Translation Layer + Retrieval + DigiKey integration + streaming UI. Validate with 2–3 paying customers. Learn what the Translation Layer is missing — and fix it.

✓ Gate: 80%+ recommendation accuracy on validated test scenarios
6 – 12 mo
Phase 2

🌐 Multi-Persona + Distributor & OEM

Add Distributor and OEM personas. Expand data integrations (Mouser, Arrow). Customer ERP connector. Customer admin console. Translation Layer refined to 500+ application patterns.

✓ Gate: A distributor deploys Peregrine for their customers
12 – 24 mo
Phase 3

⚡ Beyond Connectors — Full Passive BOM

Extend Translation Layer to resistors, capacitors, inductors, discrete semiconductors. New manufacturer API integrations. Peregrine handles BOM-level design assistance across a full bill of materials.

✓ Gate: Peregrine handles 80% of component selection across a full BOM
24+ mo
Phase 4

🚀 Platform — Self-Serve SaaS

True multi-tenant SaaS. Self-serve onboarding. Marketplace of pre-built data connectors. Partner API for distributors and OEMs to embed Peregrine in their own products and websites.

✓ Gate: ARR, net revenue retention >100%, customer count
07 — Business Model

Consulting-led. SaaS at scale.

Peregrine enters the market through paid consulting engagements — profitable from day one, building the customer relationships and data integrations that become the SaaS foundation. Every engagement funds the next capability. Every integration raises the switching cost.

Phase 1 — Consulting Engagement

  • Scoped deployment: infrastructure + data integration + onboarding
  • Typical range: $50K – $200K per engagement
  • Profitable from day one — not a loss-leader
  • Creates the data foundation for the SaaS layer
  • High switching costs built in from the start

Phase 2 — SaaS Subscription

  • Recurring revenue post-deployment
  • Scales with users, API volume, data sources
  • High-margin once platform is built
  • Expansion is natural — new personas, new categories
  • Net revenue retention >100% target
Consulting revenue is immediate — cash flow from day one. SaaS revenue is recurring — predictable and scalable. Switching costs are high — once ERP is integrated and engineers rely on Peregrine daily, the relationship compounds. This is the flywheel.
08 — Immediate Actions

From plan to product in 30 days.

The POC is built and validated. The architecture is designed. The domain knowledge is in the room. The audit is complete. The next step is a straightforward set of actions that move Peregrine from plan to Phase 1 build.

Proposed Next Steps

These five actions move Peregrine from proof-of-concept to Phase 1 build in the next 30 days.

1 Domain Knowledge Session — Mark, Bob & Kano
2 Identify Pilot Customer (End User persona)
3 Phase 0 Infrastructure Kickoff (2 weeks)
4 Ledge AI Competitive Briefing (Bob)
5 Scope Phase 1 Consulting Engagement Pricing
👥

The Domain Knowledge Session

Mark Beatty (15yr TE), Bob Mulfinger (TE career / sales), Joe Olson (TE engineer) + Kano. This is the session that seeds the Translation Layer with real connector engineering expertise that no competitor can replicate. Everything starts here.

🎯

The Pilot Customer

One OEM or distributor willing to deploy Peregrine and provide real-world feedback. Paid engagement. The goal is not perfection — it's validation and learning. Everything after that is refinement.

🏗️

Phase 0 Infrastructure

Two-week sprint: Azure infrastructure, Key Vault, Entra ID auth, LangGraph scaffold, streaming API skeleton. All security issues from the audit resolved before a single line of product code is written.

⏱️

The 18-Month Window

The distributors and OEMs are moving. DigiKey has an AI team. If we move now, we build the Translation Layer moat before anyone else gets close. That window closes. We move first.