Lloyd Clark, PhD
Founder & CEO, FDRL.SI
I build and operate software businesses across defense, aviation, and industrial services. At FDRL.SI, I lead product direction and customer delivery, and remain hands-on with software architecture and implementation.
My engineering work focuses on local inference, model evaluation, data pipelines, and constrained optimization, including systems designed for disconnected environments.
- Model evaluation: compare model output with deterministic baselines, inspect failure cases, and measure the cost of review and rework.
- Disconnected systems: package local inference, retrieval, evaluation, pinned artifacts, and explicit dependencies for operation without cloud access.
- Decision systems: combine model interpretation with deterministic checks, constraints, and approval tied to the proposed action.
I use Decision APIs for interfaces that return a recommendation together with its evidence, constraints, uncertainty, and approval requirements. Architecture note.
| Work | Engineering focus | Status |
|---|---|---|
| GoldenAye Flow Lab | Apache NiFi flow inspection, evidence-linked explanations, and ingest-policy comparisons with deterministic replay. Validation tool. | Pilot; disconnected deployment under evaluation. |
| SPLap | Interactive analysis and SPL learning using real telemetry. Built with my daughter Lilli for Splunk Enterprise and Splunk Cloud. | Published Splunk app. |
| Chalker | Aircraft scheduling proposals with maintenance conflicts, turnaround buffers, owner review, and daylight-saving-time checks. Built on Builder.io Agent-Native. | Headless prototype; no bookings. |
| Motor Quote Workbench | Motor specification screening, editable quote spreadsheets, and gross-margin pricing on Univer. | Local browser prototype; no model inference. |
Related builds: Flow Doctor research, based on Strands Harness SDK, and FDRL.SI Preflight, based on Google AX. Their repositories document my additions and validation limits. The aircraft-readiness demonstration uses a synthetic scenario.
I publish inputs, methods, raw outputs, and failure cases so another engineer can reproduce and challenge the results.
- NiFi diagnosis: test whether a local model can distinguish a configuration concern from an unsupported claim about runtime behavior, using synthetic cases, deterministic baselines, and strict output grading.
- Ingest economics: compare sampling policies at equal byte budgets across 20 seeded trials, with explicit cost assumptions and reproducible charts.
Experiment source and results · Methods and limitations
These are development experiments with synthetic data; software checks, model performance, and business assumptions are reported separately.
Recent notes cover cost per useful outcome, approval as software architecture, and evaluation without cloud access. All field notes.
I own Blue Ridge Electric Service, an industrial motor service business. Working with repair, quoting, documentation, and operating costs informs the software problems I choose.
I am a commercial pilot and flight instructor, with a PhD in Artificial Intelligence and CISSP / ISSEP certifications. Lilli and I also built No Brashers, a pilot workspace for checklist prompts and debrief notes. AOPA covered the project.
Tools: Python · SQL · Apache NiFi · Splunk · local LLMs · retrieval · evaluation · constraint optimization · MCP · simulation
Email · Speaking and bio · Aviation projects
Banner: generated concept artwork. Public examples identify synthetic data and validation limits. GoldenAye Flow Lab is independent of the Apache Software Foundation.