Any company can measurably succeed in AI.
Iron Rod Systems exists to make that true for any company with a real problem: the ones that don't know where to begin, and the ones that already tried, alone or with another partner, and didn't get there. We start with one real problem, prove it on your own data in weeks, and grow what works. Every result traces to its source, and every claim can be checked.
AI fails companies for ordinary reasons.
Not because the models are weak. Because the project started with a tool instead of a problem, because nobody agreed on what a correct result looked like, and because the first thing built was a demo rather than something that runs. We founded Iron Rod Systems to do the opposite, every time.
Engagements begin with the thing that costs you time or money today, and the one measurement that would prove it fixed.
A working system on your real data, scored against your ground truth, typically in two weeks. You see it run before you fund a build.
Most clients start with one win and grow from there. The second project reuses the first one's configuration, sources and rules.
Numbers trace to their source. Answers carry grades. Records are signed. That is what we mean by trusted AI, and it is what we built TruthLock to enforce.
Start small. Grow what works. Keep one lead.
Engagements run from a few hours of expert help to multi-year builds, and most of them grow from a small first step.
A working session on the problem and the data, systems or documents behind it. We agree on the measurement that defines success before anything is built.
A system that runs on your real data and is scored against your own ground truth. Fixed price when the scope is clear.
Tenancy, security, monitoring, failover and documentation. Deployed in your cloud or ours, reviewed before go-live.
Your team owns it, or we run it. The next problem starts from what already works, with the same accountable lead and the specialists the work needs.
Products born from real problems.
Iron Rod Systems was founded in Salt Lake City in April 2025 by Jeff Hunt, after forty-three years of building systems that had to work in production and three years leading AI product and strategy at a Salesforce partner. The company set out to do two things: consult on AI across the whole lifecycle, and build products that answer for themselves.
The first product was LDS Iron Rod AI, a scripture study tool self-funded by Jeff and his wife. Building it exposed the problem that became the company's center of gravity: when an AI gets something wrong about a subject that matters, it does not feel like a technical error. It feels like being lied to. So we built the layer that grades every claim against ranked sources and signs the record. That layer became TruthLock, now a product in its own right, with two patent applications pending.
The capabilities grew in the same order a client would need them. It started with an AI service bot that answered from a company's own documents. Retrieval was added so the answers could cite their sources. Then advanced document ingestion, so manuals and specifications came in whole. Then computer-vision ingestion, so schematics, drawings and photos could be read, not just the text around them. Then agentic AI, so the system could take steps, not only answer. Each stage was built for a real client and kept for the next one. A telemetry analyst built for a NASCAR Cup Series team became Breathing Dragon AI; document work for manufacturers became MIDAAS.
Today Iron Rod Systems delivers across all of AI: strategy, agents, retrieval, computer vision, analytics, platforms and integration, with one accountable lead and a network of specialists brought in when the work calls for them.
- Founded
- April 2025, Salt Lake City, Utah
- What we do
- AI consulting from hours to years, and products we build and run
- Live products
- TruthLock · LDS Iron Rod AI
- In development
- MIDAAS · Breathing Dragon AI
- Patents
- Two pending, on TruthLock verification and proof of authenticity
- How to start
- One problem, one measurement, a proof in about two weeks. Contact us.
Five commitments that don't change with the client.
Every engagement has a number agreed up front, and we report against it, including when it falls short.
Verified means exact, in scope and current. Everything else is labeled with the reason it fell short, and refusal is a legitimate state, never hidden.
Deployed in your cloud when you want it there, never used to train anything, never held hostage at handover.
We explain what we built, what it costs, and what it cannot do, in words a decision-maker can act on.
Our own site runs on our own agent, verified by our own product, and reviewed with the same security standards we apply for clients.
Capabilities
What Iron Rod Systems delivers. When an engagement calls for specialized skills, we bring in proven specialists from our network, under one lead and one contract. Not a keyword list; a working inventory.
Python, TypeScript, JavaScript, Apex, SOQL and SOSL, SQL, Cypher, HTML and CSS, Bash, YAML and Terraform HCL, Java and PHP historically, C and assembly at the start.
Anthropic Claude (Opus, Sonnet, Haiku), OpenAI GPT-4 through GPT-5, Google Gemini, Azure OpenAI, AWS Bedrock, xAI Grok, Groq, DeepSeek, Hugging Face, Nvidia and Snowflake Arctic, LlamaCloud, Qwen vision models, Cohere and Pinecone rerankers, OpenAI and Google embeddings.
Retrieval-augmented generation at scale, including index dilution fixes; multi-agent orchestration; ReAct tool-calling agents; text-to-SQL with self-correction; prompt registries with versioning; config-driven presets; guardrails for injection, PII and length; evaluation harnesses with golden datasets; LLM routing with failover and circuit breakers.
LangGraph, LangChain, LiteLLM, Vanna, Vercel AI SDK, NeMo Guardrails, RAGAS, Langfuse, Docling, PyMuPDF, RapidOCR and Tesseract, LlamaParse, AWS Textract, Pinecone Assistant, Model Context Protocol servers and clients, Composio connectors.
Show the full inventory, 15 more categoriesShow less
RF-DETR and Roboflow, Segment Anything, vision-language models for schematics and photos, OpenCV, legend-template matching, tiled COCO datasets, active-learning loops, training on CUDA, Apple silicon and rented GPUs, symbol calibration and readiness grading.
Claude Code, Cursor, v0, ChatGPT and Grok as daily tools. Platforms designed so AI coding tools produce tenant-isolated, secure applications by construction, with rules, skills and compliance ledgers enforced in the repo.
ECS Fargate, EC2, RDS PostgreSQL, S3, Lambda, Bedrock and Bedrock Knowledge Bases, Textract, OpenSearch Serverless, KMS, Secrets Manager, SES, SNS, CloudWatch metrics, alarms, dashboards and Synthetics canaries, Systems Manager, PrivateLink, IAM, all provisioned with Terraform.
Gemini models and Google AI Studio, Google Cloud Run, Google Drive and Docs APIs, Google identity and OIDC sign-in, Google Play and Workspace integration.
Azure OpenAI, Azure AI Search, Azure AD and Entra identity, Microsoft 365 and SharePoint as document sources, ServiceNow and Office integrations, Windows and .NET platform integration work.
Cloudflare Workers, Pages, Tunnel, WAF and Turnstile; Vercel; Railway; Render; Neon serverless Postgres; MotherDuck; Neo4j Aura; Pinecone; Weaviate; vast.ai GPU rental; Docker and Nixpacks on any of them.
Sales Cloud, Service Cloud, Field Service, Manufacturing Cloud, Revenue Cloud, Marketing Cloud, Financial Services Cloud, Data Cloud, Experience Cloud, Einstein, Bots, Agentforce, Flow, Apex, Lightning Web Components, managed packages, Bulk API 2.0, JWT bearer OAuth, Data Loader, SFDX. Ten years as a Principal Solution Engineer, with Field Service, Data Cloud and Agentforce implementation partners for full rollouts.
AI wired into the Now Platform: cited answers and agents inside ITSM, CSM and the service portal, knowledge and incident workflows, REST and Table API integrations, IntegrationHub, Flow Designer, scoped applications and Now Assist. Delivered with ServiceNow platform developers and integration specialists from our network, under one Iron Rod Systems lead.
PostgreSQL with pgvector and row-level security, TimescaleDB, DuckDB, Neo4j graph, Weaviate, Pinecone, MongoDB, OpenSearch, Azure AI Search, Redis, MinIO and S3 object storage, Snowflake, SQLite, Oracle, Alembic and Prisma migrations, dlt pipelines, Parquet.
REST and GraphQL APIs, MCP, webhooks with HMAC signing, Salesforce, ServiceNow, Jira, HubSpot, Slack, NetSuite, QuickBooks, Workday, SharePoint, Google Drive, DocuSign CLM, ServiceMax, Stripe metered billing, RevenueCat, Resend, OneSignal, Calendly, Formspree.
React 19, Vite, Next.js, Tailwind, shadcn/ui, Recharts, TanStack Table, OpenSeadragon deep zoom, FastAPI, Flask, Jinja2, Nginx, SSE streaming, accessibility with ARIA, iOS apps via Median and Capacitor, App Store submission, Apple and Google sign-in.
OWASP ASVS and API Top 10 reviews, JWT and API-key auth, RBAC with per-tenant scoping, SSO with Google, Okta and Azure AD, TOTP-gated deploys, zero-trust tunnels, KMS and Ed25519 signing, SHA-256 Merkle attestation, blockchain anchoring on Base, audit logs, secrets management, PII redaction.
Docker Compose, Terraform, GitHub Actions CI, semver releases with checksums, Sentry, Prometheus and OpenTelemetry, Slack alerting, health checks and canaries, pytest, Vitest, headless Chromium end-to-end tests, structured logging, runbooks and in-app documentation.
Oracle database, CRM and business intelligence; ERP implementation across finance, purchasing, inventory, shop floor and reverse logistics; CPQ and quote-to-cash; contract lifecycle management; field service management; Unix systems programming.
Manufacturing, construction and electrical contracting, lighting controls, HVAC, utilities and power products, rail, industrial equipment, professional services, motorsports, financial services, life sciences, faith-based consumer software, small business.
Jeff Hunt, founder and CEO
Forty-three years of systems that had to work in production, from Unix kernels in the 1980s to a decade at Oracle, a decade as a Principal Solution Engineer at Salesforce, and three years leading AI product and strategy at Uptima, where he invented Elle. He founded Iron Rod Systems in 2025 and is the inventor on the two pending TruthLock patents, held by the company.
