Former FBI senior intelligence analyst and supervisor (24 years). Now building data tools through AI-assisted development. Houston, Texas · LinkedIn · mikejennings.dev
I spent my career turning large, messy, multi-source data into decisions for investigators, prosecutors, and senior leaders. In 2026 I started building the tools myself.
I direct AI coding assistants (Claude and GPT-family models) to write the code. I own the problem definition, the data model, the architecture, verification, and every release decision. These projects use the Elite Dangerous exploration community's open galaxy data as a live sandbox: a 100+ GB dataset covering more than 200 million star systems, with real users and no sensitive data.
- Specification first. Requirements and design notes are written before any code is generated.
- Guardrails. Documented architecture rules that every assistant has to follow.
- Independent review. Code from one model is reviewed adversarially by another, then by me, and checked by tests before it is released.
- Validate against authoritative sources. Outputs are confirmed against external data, never accepted on a model's say-so.
- Human in the loop by design. I piloted an autonomous multi-agent workflow, measured its real coordination and verification costs, and went back to a tighter human-in-the-loop model.
- Lessons learned. A written record of what went wrong and what changed because of it.
| Project | What it does | Analytic pattern |
|---|---|---|
| Monops: Entity Screening Toolkit | Screens researchers and organizations for foreign affiliations by reconciling declared ties against open sanctions, corporate ownership, grant, and bibliometric data, with explainable scoring and an analyst reviewing every match. Live demo | Entity resolution and due-diligence screening (research security / KYC) |
| ED Sector Surveyor | Streams a 113 GB compressed dataset into per-sector SQLite databases, infers likely-but-unrecorded systems from naming-pattern gaps, and keeps only candidates an external catalog confirms are new | Large-scale streaming ETL; generate-then-validate lead generation against an authoritative source |
| ED Colony Scout | Ranks targets with an editable scoring rubric and tests reachability as multi-hop jump chains | Explainable, adjustable scoring; network reachability analysis |
| ED Expedition Ledger | Ingests live event logs into SQLite with hash-based de-duplication and exports per-leg Excel and CSV reports | Idempotent, auditable ingestion; structured reporting from raw event data |
| Ring Density Monitor | Compares each new observation to a population baseline in real time and flags outliers | Real-time monitoring and anomaly flagging |
Python · SQLite · Tkinter · PyInstaller / Inno Setup · GitHub Actions · unittest / pytest · Excel and Power BI for analysis
The Elite Dangerous tools are unofficial fan projects, not affiliated with Frontier Developments.