State
Typed requests, explicit provider priority, validated output contracts, and review decisions.
Agentic Systems
Reliability mechanics you can inspect.
I bring production discipline from computer vision and edge deployment into agent and multimodal workflows. The public evidence below separates tested orchestration code from deterministic browser applications and private production experience.
Failure handling
Transient retry, permanent fallback, contract rejection, confidence gates, and preserved traces.
Evaluation
Executable hard cases for the primary path, retries, fallback, invalid outputs, review gates, and total failure.
Reliable Agent Workflow reference
A dependency-free Python implementation of the orchestration layer, with ten executable tests.
System shape
A small provider protocol keeps model adapters outside the workflow core. The router records every attempt, retries expected transient failures, falls back on permanent or contract failures, validates structured outputs, and forces human review below the request confidence threshold.
request -> route(primary) -> attempt -> transient failure
-> route(fallback) -> validate contract
-> confidence gate -> accepted output + trace
Executable evidence
The hard-case suite covers primary success, retry recovery, invalid-contract fallback, confidence-based human review, total provider failure, and evaluation reporting.
Photography Mentor with browser-local retrieval
A browser-local multimodal learning application that joins image diagnostics, retrieval, and actionable practice guidance.
System shape
Use local photography knowledge retrieval together with histogram, brightness, color, sharpness, composition, and EXIF analysis. Keep the evidence path visible and translate observations into specific shooting or editing actions.
Engineering signal
Multimodal input handling, local retrieval, deterministic image analysis, constrained coaching output, review state, and privacy-aware browser execution. It is intentionally described as deterministic rather than as an LLM-backed agent.
Production discipline carried forward
The differentiator is not claiming a larger model stack; it is applying deployment and evaluation discipline to a new workflow class.
My production experience is in computer vision, computational photography, edge deployment, native integration, visual debugging, and release evaluation. The public agent reference makes the adjacent orchestration mechanics inspectable without exposing employer code or presenting scripted adapters as live provider integrations.
Role fit
Applied AI and Forward Deployed Engineering where perception, workflow design, runtime constraints, evaluation, and handoff matter together.
The public evidence demonstrates provider-neutral orchestration mechanics and deterministic multimodal tooling. It complements production depth in computer vision, computational photography, edge deployment, native integration, visual debugging, and release evaluation.