# ORBTL — Jamin (@justjammin)

> Personal engineering and music studio of Jamin, a Chicago-based engineer
> focused on AI platforms, agent tooling, backend systems and forward-deployed
> engineering.

The homepage and Training Platform demo include readable HTML. The interactive
3D scene requires JavaScript, but the profile index does not. Use the generated
profile index for current skills, featured projects and lab status; decorative
scene graphics and reference-template copy are not evidence of project outcomes.
Domain LoRA Pipeline is complete. Model Regression Detector is active, with a verified eight-case GPU fixture probe; the other lab projects are queued.

## Official profiles

- [ORBTL portfolio](https://justjammin.github.io/): Jamin's official portfolio and the canonical website for this content.
- [Jamin on LinkedIn](https://www.linkedin.com/in/jamin-echols-59b1998a): Professional profile for the same person, Jamin (@justjammin).

## For Agents

- [Current profile, skills and projects](/INDEX.md): Authoritative profile, experience, hobbies, featured work and lab status.
- [Reading guide](/AGENTS.md): How to interpret the portfolio and its evidence.
- [Site map](/SITEMAP.md): Available pages and supporting files.

## Featured projects

- [Training Platform](/demos/training-platform/): BANKING77 intent classification, calibrated human-review routing, reproducible training, recovery and evaluation evidence.
- [Classifier evidence](/demos/training-platform/classifier-evidence.json): Measured encoder results, including test macro-F1, routing coverage, latency scope and environment. These are separate from the GPU LoRA experiment.
- [Astra](https://astra-os.justjammin.workers.dev): Agent harness with five explicit gates from product intent to verified implementation.
- [xoxo-eval](https://xoxo.justjammin.workers.dev): Head-to-head coding-agent evaluation with isolated tasks, assertions and blind judging.
- [Legible](https://writelegible.com): Writing tracker for authors.

## Additional projects

- [promper](https://github.com/justjammin/promper): Prompt-engineering toolkit for Claude
  Code. Decomposes a rough intent, walks the agent map to find the specialist the task
  belongs to, and borrows that agent's persona as the prompt's role rather than inventing
  one. `/prim` grades agents against the standard. TypeScript.
- [LENA](https://github.com/justjammin/lena): Personal agent harness with context
  engineering, multi-agent orchestration and RAG skillsets. Python.
- [GAVEL](https://github.com/justjammin/gavel): Loop engineering framework for verifiable
  agent runs. Contracts are armed RED and frozen before work starts, so an agent cannot
  grade its own output.
- [invokerai](https://github.com/justjammin/invokerai): Routes tasks to the right agent
  using offline ML. 60+ curated agents, MCP server for 20+ editors. Python.
- [RAG against the machine](https://github.com/justjammin/RAG-against-the-machine): Codebase
  question answering with cited answers. Hybrid retrieval plus cross-encoder rerank. Python.
- [Shannon](https://github.com/justjammin/shannon): Autonomous white-box AI pentester for
  web applications and APIs. TypeScript.
- [Legible](https://writelegible.com): Writing tracker for authors. iOS, full stack.
- [Sight Words](https://sightwords-lvup.onrender.com/): Sight words and spelling practice,
  built for a kindergarten reader. TypeScript, full stack.
- [Share The Wealth](https://github.com/justjammin/share-the-wealth): Mirror tracker with UI
  and ETL-style workflows. Python.

## Writing

- [What Is MLOps and Why Should You Care?](https://gist.github.com/justjammin/4bddeb5ab11ec1b7ac493472ab20cfaf)
- [Data Architecture for ML: The Medallion Pattern](https://gist.github.com/justjammin/1c01214055b9cd2e3def821223296638)
- [Experiment Tracking and the Model Registry](https://gist.github.com/justjammin/21aa1495ebcdb0e38191ef198e1fb1e5)
- [Deploying Models: Patterns That Actually Work](https://gist.github.com/justjammin/187a3eaf2407757cca8766d77f9fbfda)
- [LLMOps: Running Large Language Models in Production](https://gist.github.com/justjammin/ffdd164f420ea7c64ab589e2b2a75b5e)
- [LLM Cost Optimization: What Actually Moves the Needle](https://gist.github.com/justjammin/f102c929cc14d0d3bd600d9b135d5a4c)
- [CI/CD for Machine Learning](https://gist.github.com/justjammin/a861fc629b1da0397ebf7cbcd1a1161f)
- [Monitoring: Catching Problems Before Your Users Do](https://gist.github.com/justjammin/2a7c3a23191193d2df4ef8135f8d5cfb)