Jeremy Cleland · Agentic AI
Agentic AI & Intelligence SystemsI build AI-enabled software for complex, real-world problems, with a focus on agentic AI and intelligence systems.
Engineering with context
Hands-on AI engineering. An M.S. in Artificial Intelligence. A career shaped by service as a Special Forces Medic.
At Gray Matter Resources, I bring hands-on software engineering, applied AI, and Special Operations experience to problems where the quality of information matters.
An architectural diagram of an agentic system. A dense network of retrieval, tools, evaluation, and routing simplifies into one workflow that acts on its own and escalates to human judgment when the decision is consequential.
01 / Selected work
Built a local incident investigator with real MCP tool calls, evidence-linked proposals, approval-gated simulation, and inspectable recorded evaluations.
class IncidentAgent:
def evaluate(self, trace):
return eval_metric(trace.evidence)Built multi-agent and tool-calling workflows for clinical evidence and veteran benefits, with human review for consequential decisions.
workflow = (
Graph()
.add_node("extractor", clinical_llm)
.add_edge("human_review")
)Developed a plant-disease classifier using an attention-augmented ResNet18 and inspected model behavior with evaluation visualizations.
logits = cbam(resnet18(x))
pred = logits.argmax(dim=1)02 / Recent writing
Sep 21, 2026 · Artificial Intelligence
Most cycle time in complex work is coordination—not the task. Agentic AI pays off when one recurring workflow is redesigned around that tax, with human judgment where accountability sits.
Read essay : Cutting the Coordination Tax: Redesign One Workflow FirstFort Bragg uses analytics to reclaim training hours and spends them on tourniquet conversion — unassisted judgment. The documented effect of assistive AI is that unassisted judgment decays. The January 2026 Department of War AI Strategy pulls the other way.
Deterministic tool tests, trajectory invariants, and outcome checks against observable state. A working three-layer evaluation stack, shown through a public MCP agent repo, including the parts that did not work.
03 / Learn
Comprehensive guide for choosing the right algorithm based on problem constraints, data characteristics, and performance requirements.
Complete reference for asymptotic notation, complexity analysis techniques, and time/space complexity bounds for algorithms and data structures.
Master fundamental concepts of algorithm analysis including Big O notation, recurrence relations, Master Theorem, and amortized analysis.
Master backtracking algorithms including N-Queens, sudoku solving, subset generation, and constraint satisfaction problems.
04 / Craft
Agents, pipelines, and evaluation harnesses.
Incident Lab and production claims workflows.
Multi-agent graphs with human review points.
Recorded traces, test cases, and failure inspection.
PlantDoc, HMER, and other applied-ML work.
Service interfaces around retrieval and tools.
Evidence stores and evaluation datasets.
Reproducible training and inference environments.
Path
Current
Bringing hands-on AI engineering, software development, and U.S. Army Special Operations experience to Gray Matter Resources.
August 2025 – May 2026
Developed and shipped prototypes and production agentic systems for veteran disability-benefits workflows in two-to-four-week cycles.
2023 – 2025
University of Michigan-Dearborn. Graduate research on applied ML pipelines.
2014 – 2022
U.S. Army 7th Special Forces Group (Airborne). Two deployments to Colombia and two deployments to Afghanistan.
05 / Engineering approach
Work with domain experts to define the task, the evidence, and where human judgment belongs.
Build retrieval and tool-calling workflows with clear interfaces and deliberate review points.
Use repeatable test cases, inspect failures, and iterate on reliability through actual use.
06 / Background
I am Director - Senior Software Engineer at Gray Matter Resources. My background includes agentic systems for clinical-evidence and benefits workflows at VetClaims.ai, and ML pipelines as a graduate research assistant at the University of Michigan-Dearborn.
I completed my M.S. in Artificial Intelligence in 2025. I retired from the U.S. Army in 2022 after serving as a Special Forces Medical Sergeant.
More about my backgroundLet’s connect
Based in Michigan. Connect about agentic AI, software engineering, and the people using what we build.