Craig Ramos Software / AI / Systems

Craig Ramos / Software + AI + Systems Engineering

I BUILD SYSTEMS THAT HAVE TO ANSWER TO REALITY.

Warehouse floors, agent evidence, spatial interfaces, and operating systems. I work where software has to survive contact with the physical world.

Five systems, one line of work
First-person view inside the modeled Plant 076 production floor with a forklift, pallet load, safety rails, and equipment Matched warehouse camera before local inventory data is visualized, showing the modeled storage structure largely unpopulated Same warehouse camera after supported local inventory data is visualized as pallet loads throughout the modeled facility

01 / PPK076

I MODEL PHYSICAL SYSTEMS.

Physical operations, modeled in software.

  1. I model physical systems. Physical operations, modeled in software.
  2. I turn operations into decision systems. 176 pallet positions recovered. 22 storage bins freed.
  3. I test what agents actually do. Agents tested against preserved evidence.
  4. I rethink how the computer can feel. Spatial computing, persistent by design.
  5. I build below the application layer. A from-scratch, verification-driven operating system.
  6. Physical systems. Software systems. AI systems. Computer systems. I build where those layers meet.
Six authored evidence states spanning physical operations, decision systems, agent evaluation, spatial computing, and low-level systems.

The line through the work

I learned systems through materials, machines, failure, movement, and consequence before I learned to express them in code.

That history changes how I build. State has to correspond to something. Authority has to belong somewhere. Evidence has to survive the demo.

Selected work / 2019–2026

Five systems. Five different kinds of proof.

01 / PPK076

Model the place before abstracting the work.

A warehouse simulation built from actual industrial geometry, then connected to supported local data for navigation, training, and spatial review.

Open case study
First-person view inside the modeled Plant 076 production floor with a forklift, pallet load, safety rails, and equipment
Plant 076 modeled as an operational environment, not a decorative scene.
Matched warehouse camera before local inventory data is visualized, showing the modeled storage structure largely unpopulated
01 / Structure
Same warehouse camera after supported local inventory data is visualized as pallet loads throughout the modeled facility
02 / Operational state

02 / Warehouse optimization

176pallet positions recovered
22storage bins freed

Measured result. Human-verified authority. No SAP write-back.

Operational rules made executable.

Consolidation recommendations respect material separation, location types, side-bin constraints, and the gap between exported system state and physical reality.

Open case study

03 / Skill Evaluation Lab

A result does not count unless its evidence survives.

Controlled original, no-skill, and candidate runs. Preserved failures. Exact artifacts. Cross-model replication. Claims narrowed when effects do not reproduce.

Open case study

evidence / experiment-0031

01 ORIGINALbaseline captured
02 NO-SKILLcontrol isolated
03 CANDIDATEruntime fixed
04 FAILUREdirty state preserved
05 VERIFYhashes compared

EVIDENCE BEFORE CLAIM

Saved M2A room checkpoint with a spatial screen placeholder, table, brick objects, object panel, trusted controls, and connected state

04 / Workspace Environment vNext

The desktop becomes a place.

Persistent objects, spatial software surfaces, durable identity, and explicit host authority inside a navigable room.

Current evidence: the room and object system are real. The large application screen shown here remains a placeholder at this checkpoint.

Open case study

05 / PythOS

Build below the layer everyone else assumes.

A from-scratch x86-64 operating system built around persistent objects, explicit capabilities, recovery, and physical verification.

Names locate. Relationships describe. Capabilities authorize.
Open case study
PythOS evidence terminal running on a physical laptop, with enough screen bezel visible to establish the hardware context
313 verification markers / zero drops / CRC 176F4C6E
Target-specific physical evidence.

Evidence context

VERIFIED WORK

The credential supports the evidence. It never replaces it.

Operational systems
176 pallet positions recovered / 22 storage bins freed
Physical systems
313 verification markers / zero drops / CRC 176F4C6E / target-specific physical evidence
Professional practice
React, Node, Git/PR workflows, testing, agent SKILL development, documentation, handoff

Background

The software came after the systems.

  1. 01Mechanical

    Parts have tolerances.

  2. 02Structural

    Systems carry loads.

  3. 03Electronics

    Failures propagate.

  4. 04Warehousing

    Software state and physical state diverge.

  5. 05Software

    Those constraints become executable systems.

I am building toward software that understands more than screens: objects, space, authority, persistent state, hardware, and the evidence required to trust an action.

Craig Ramos / Louisiana

Let the work answer first.

EmailGitHubLinkedInSource