Vertical AI Architecture

18+ specialized AI agents powered by Anthropic Claude, backed by free federal CMS data that competitors charge six figures to access. Purpose-built for medical device sales intelligence.

18+ Agents

Anthropic Claude

Primary AI model for reasoning, analysis, and content generation across the entire agent fleet

14K+ Prospects

CMS Data Moat

Free, Congressionally mandated NPI registry and Open Payments data — the dataset competitors sell for $100K+

2 Score Engines

Dual-Engine Scoring

Every practice scored on both Readiness and Opportunity — two independent signals that together identify the best targets

Multi-Agent Architecture

Specialized AI agents handle discrete tasks — discovery, enrichment, scoring, and outreach — coordinated into a single intelligence pipeline.

Anthropic Claude

Primary AI Engine

Powers all reasoning, analysis, scoring, and content generation across the agent fleet. Handles prospect research, sales brief generation, competitive analysis, and enrichment workflows.

Key Capabilities:

Prospect intelligence extraction
Sales brief generation
Device & service identification
Competitive landscape analysis

OpenAI Routing Layer

Agent Orchestration

Lightweight routing layer that directs queries to the right specialized agent. Handles task classification and workflow dispatch across the 18+ agent fleet.

Key Capabilities:

Task classification
Agent dispatch
Workflow routing
Cost-efficient delegation

CMS Federal Data Engine

Data Intelligence

Integrates free, Congressionally mandated datasets — NPI registry for physician identification and Open Payments for manufacturer-to-physician financial relationships. This is the data moat.

Key Capabilities:

NPI physician registry
Open Payments financial data
Practice geolocation
Taxonomy classification

How They Work Together

Practice Discovery

Multi-source identification using CMS NPI data, geospatial mapping, and web intelligence

Deep Enrichment

18+ AI agents extract staff, services, devices, social signals, and competitive positioning

Dual-Engine Scoring

Readiness + Opportunity scores per practice, weighted by specialty and territory

Sales Brief Delivery

Actionable one-pagers with practice intel, device inventory, and recommended talk tracks

Technology Stack

Built on modern, proven technologies. No unnecessary abstractions — just the right tools for each job.

Frontend

Modern web stack with server components, strict TypeScript, and the Bun runtime for fast builds and execution

Next.js 15React 19TypeScriptTailwind CSSBun

Data & Storage

Supabase PostgreSQL with row-level security, real-time subscriptions, and built-in storage for documents and media

SupabasePostgreSQLSupabase StorageRow-Level Security

Infrastructure

AI agents deployed as Docker containers on Railway. Dashboard hosted on Vercel with automatic deploys from Git

RailwayDockerVercelCloudflare

Security

Encrypted in transit and at rest, database-level access controls, and separate sandbox and production environments

TLS 1.3AES-256Row-Level SecurityEnvironment Isolation

Integration

Direct integration with federal CMS datasets, plus APIs for CRM connectivity and geospatial practice mapping

CMS NPPESOpen PaymentsREST APIsWebhooksGeospatial APIs

AI & Agents

18+ specialized Python agents for prospect research, scoring, enrichment, and outreach — all orchestrated through a central event bus

Anthropic ClaudeOpenAIPythonFastAPI18+ Agents

Built for the Job

18+

Specialized AI Agents

14K+

Prospects Enriched

2

Independent Score Engines

Every component chosen for a reason. Supabase for real-time data, Railway for isolated agent deployment, Vercel for instant frontend deploys.

What Makes It Different

Not another AI wrapper. Fathom combines proprietary agent orchestration with federal data sources to build intelligence that didn't exist before.

Agent Fleet Architecture

18+ Python agents each handle a single job — prospect discovery, enrichment, scoring, outreach, competitive analysis. Deployed as isolated Docker containers on Railway.

One agent, one job

Dual-Engine Scoring

Every practice gets two independent scores: Readiness (how ready they are to buy) and Opportunity (how much revenue potential exists). Together they surface the highest-value targets.

Two signals, one ranking

CMS Data Integration

Free federal datasets that competitors charge $100K+ for. NPI registry identifies every physician. Open Payments reveals manufacturer-to-physician financial relationships.

Free data, paid-tier intelligence

Deep Practice Extraction

AI agents extract device inventories, staff rosters, service menus, social signals, and business indicators from practice websites — building 16-category intelligence profiles.

14K+ practices profiled

See the Architecture in Action

Walk through how our agents discover, enrich, score, and deliver intelligence on medical practices — in real time.

Schedule a Demo