Ohmni Oracle

Electrical Field Intelligence

Collin VelardeFounder + BuilderField-Tested Systems

Builder Branch

Field-tested judgment. Serious systems.

I'm Collin Velarde, a General Foreman commercial electrician turned software builder. I build field intelligence, retrieval infrastructure, ETL pipelines, and private operator systems that can hold up under real use. Ohmni Oracle is the clearest public proof, not the limit of the work.

See The Work Request A Walkthrough

The theme is simple: the next edge belongs to domain experts who can build with AI instead of handing the whole problem off to somebody who only knows the tooling.

Working Thesis

Good domain judgment should not die at the handoff.

The job is to make hard-won operator knowledge legible, usable, and fast without flattening it into startup copy or prompt tricks.

Product behavior tied to real workflow pressure

Retrieval and ETL built as system layers, not afterthoughts

Private proof kept private when trust matters

20+ years in commercial electrical

6-repo Ohmni Oracle program

Live product + retrieval + ETL

Private workflow systems

FounderGeneral ForemanSystem Builder

Who I Am

I came into software from the field, not from a pitch deck.

I spent more than two decades in commercial electrical work, including General Foreman responsibility in the Chicago market. That background is not a brand accessory. It is the reason the software starts from sequence, coordination, drawings, code, failure modes, and the people carrying the job when it gets tight.

When I build, I do not start with model novelty. I start with the burden, the workflow, and the point where a bad answer or weak handoff turns into lost time, bad decisions, or avoidable cleanup.

I think the future belongs to domain experts who can build with AI instead of waiting on a handoff between someone who knows the work and someone who only knows the tooling.

Field judgment first

I start with the operator burden, the failure mode, and the workflow that actually costs money when it breaks.

Production over demo

Routing, validation, access control, smoke checks, and safe fallbacks matter as much as the model call.

Systems, not prompt theater

I build the retrieval layer, ETL, product surface, and operating seams that make AI usable under pressure.

Range without drift

Ohmni Oracle is the clearest public proof, but the underlying skill set extends well beyond one electrical app.

Selected WorkShipped ProofPublic-Safe Receipts

What I Build

One domain gave the work its edge. The systems work goes much wider than one app.

The public proof has to stay honest. That means showing what has actually shipped, what is safe to show, and what kind of systems thinking sits underneath the product surface. The point is not to flex a stack. The point is to make judgment legible through receipts.

01 | Public Product

Ohmni Oracle

Field intelligence for electrical teams working from drawings, code, project context, and messy field input.

Shipped punch-list export from rough walkthrough notes into a structured workbook crews can use.

Hardened GPT-5 routing across instant, brain, and pro lanes with regression-tested model operations.

Locked down the public surface with CSP, exact OAuth origin checks, and live verification on the production domain.

This is the clearest public proof that I can take domain pain, turn it into product behavior, and harden it for real use.

02 | Retrieval Infrastructure

Routing and retrieval systems

Middleware that routes across code, definitions, tables, and field knowledge without turning into brittle magic.

Shipped schema-drift fallback for tables retrieval and kept production probes green.

Extended multi-index routing with shared domain configuration instead of ad hoc branch growth.

Validated the live retrieval layer with production smoke, keyed probes, and telemetry checks.

This is the part of the work that proves I can build infrastructure, not just a polished interface.

03 | Data + ETL

Project-agnostic ETL and knowledge pipelines

Extraction and indexing systems designed to stay reusable across projects instead of collapsing into one-off hacks.

Merged packet-pure ETL hardening on main and parity-validated it across multiple project packets.

Shipped split-preflight and title-block hardening to keep bad inputs from poisoning downstream work.

Structured a 50k+ document proprietary knowledge corpus with clear boundaries between retrieval and ETL.

This is where the serious systems work shows up: clean seams, reusable behavior, and validation before confidence.

04 | Private Client Work

Private operator systems outside the electrical lane

The same operating discipline applies outside construction, but public proof stays redacted when client trust matters more than marketing.

Built hunt-first workflow software for a real-estate operator instead of another generic CRM.

Kept evidence-first review and apply flows at the center of the product behavior.

Use public proof when it is safe and private proof when the work is sensitive.

Range matters, but judgment matters more. I will not burn a client just to make a landing page louder.

Ohmni OraclePublic ProofNot The Ceiling

How Ohmni Fits

Ohmni Oracle is the clearest public proof of how I build.

It is the public product surface where the field-first thesis, the software work, and the infrastructure work all meet. It is also only one part of the broader system stack.

If the homepage shows the product, this page should show the builder and the layers underneath it.

Frontend

Public site and operator surface

The public shell, the authenticated app, request access, and the product UI that crews and operators actually touch.

Backend

Product logic and workflow delivery

Chat, export flows, email intelligence, model routing, and the contracts that turn AI behavior into product behavior.

Retrieval

Search and routing middleware

The layer that decides where a question belongs and keeps grounded answers tied to the right surface.

ETL

Drawing and project-processing pipelines

Reusable extraction systems that keep project context structured before it ever reaches the chat layer.

Knowledge

Proprietary corpus and doctrine

A long-tail electrical knowledge layer designed as serious operating IP, not content sludge.

GitHubRepo SurfacesShipped Layers

GitHub And Shipped Surfaces

The code is part of the public proof.

The repos below map to the public product surface, the retrieval layer, the ETL stack, and the knowledge system. They are not the whole story, but they are the most direct public line into how the work is built.

View GitHub Profile

Ohmni frontend

Public site + product UI

Field-first marketing surface, request-access flow, launch hardening, and the app UI that carries the product.

Ohmni backend

Product API

Punch-list export, model routing, workflow orchestration, and the live product logic behind Ohmni Oracle.

Azure functions

Retrieval middleware

Definitions/tables expansion, safe retrieval fallbacks, and production-validated multi-index routing.

Ohmni Oracle v3

Drawing ETL

Project-agnostic packet processing, title-block hardening, and parity-validated extraction pipelines.

Super Electrician KB

Knowledge system

Corpus structure, faceting, and the knowledge boundaries that make retrieval usable instead of noisy.

Private WalkthroughWorkflow AuditShipped Systems

Start The Conversation

Start with the workflow, not the pitch.

If you want a private walkthrough, a workflow audit, or help scoping an internal tool, use this form. If the fit is outside electrical, say that clearly in the message.

Use this for a private walkthrough, workflow audit, or a serious internal-tool conversation.

If the best fit is ETL, retrieval, or system hardening instead of the app, say that directly.

If the public proof is not enough, I can walk through the private proof live where it is safe to do it.