enterprise terrain monitored through live cyber dashboards and tracking workflows.
TS/SCI cleared | CompTIA Security+ | Platform security, AI infrastructure, and inference systems
Building secure AI infrastructure for high-stakes environments.
Cleared cyber operations specialist focused on platform security software, secure model access, LLM serving reliability, production ML observability, and mission decision systems. My strongest work turns operational pressure into tested backend services, routing and policy controls, audit evidence, benchmarks, dashboards, and runbooks.
- 26 wings
- enterprise cyber support
- 216K
- tracked data points
- 30K
- assets hardened
- 2025
- A6 Airman of the Year
Mission Impact
Operational outcomes worth scanning first.
threat-resolution cycle reduction after dashboard and tracking-tool engineering.
urgent change execution improvement during combat-readiness exercise support.
concurrent players supported on Linux-hosted game infrastructure and community systems.
Proof Snapshot
Concrete systems work behind the positioning.
Inference Runtime Validation
Rust scheduler core with continuous batching, paged KV-cache admission, deterministic replay, workload-pressure replay summaries, replay capacity envelopes, prefill/decode utilization, KV-page occupancy, backend-scoped numeric tolerance, vLLM/SGLang mirror normalization, segmented release reports, streaming-token event ingestion, route and scheduler provenance coverage, TTFT and decode-token p95 checks, KV memory-pressure reporting, model-version transitions, token-trace fingerprints, structured release-triage owner hints, and promote, hold, and rollback gates.
GPU Performance Evidence
Triton RMSNorm and SwiGLU kernels with FP32 oracles, launch autotuning, cache-controlled CUDA-event measurement, raw p50/p95/p99 timing artifacts, and torch.compile baselines.
AI Reliability / Operations
Latency, throughput, failures, token/GPU-hour capacity, cost-to-serve estimates, Prometheus output, batch-invariance gates, deterministic trace artifacts, and Kubernetes-shaped benchmark execution.
Secure Mission Delivery
TS/SCI operational context combined with authenticated model access, RBAC, policy controls, request and token-budget limits, audit trails, sanitized trace export, synthetic capacity/cost planning, Grafana and Prometheus observability evidence, incident runbooks, and public-safe mission decision software.
Positioning
Where the portfolio is pointed.
AI Infrastructure / Reliability
Model-serving gateways, inference benchmarks, latency percentiles, Prometheus artifacts, routing, scheduling, autoscaling concepts, regression checks, and Kubernetes-oriented deployment patterns.
Inference Runtime Performance
Fused RMSNorm and SwiGLU kernels, launch autotuning, cache-controlled measurement, request routing, batching and caching concepts, workload-pressure replay, replay capacity envelopes, prefill/decode utilization, KV-page occupancy, instrumentation/profiling direction, memory-pressure checks, streaming-token trace validation, route/scheduler provenance, model-version transition metadata, token-path telemetry, and tail-latency investigation.
Hardware / Software Co-Design
ONNX graph partitioning, explicit accelerator and memory cost models, analog non-ideality simulation, synthesizable SystemVerilog, FPGA flows, and model-level quality tradeoff analysis.
Platform / Security Software
Authenticated service boundaries, RBAC, policy enforcement, audit evidence, request and token-budget limits, threat models, incident runbooks, AI observability, sanitized trace evidence, synthetic capacity/cost planning, Grafana dashboard provisioning, secure workload controls, and policy-as-code direction.
Forward-Deployed Mission Engineering
Cleared operational context, stakeholder translation, full-stack prototypes, observable systems, and decision surfaces that help users act under uncertainty.
Role Fit
High-value engineering work mapped to shipped evidence.
Compute / Inference Infrastructure
Triton-style benchmarking, latency and throughput reporting, failure accounting, token and GPU-hour capacity metrics, cost-to-serve estimates, Prometheus export, cluster-run job shapes, queueing/batching concepts, customer-facing serving API posture, routing/scheduling/autoscaling concepts, accelerator-aware release gates, traceable workload artifacts, and hardware-aware debugging direction.
Inference Runtime Performance
Rust continuous batching, paged KV admission, deterministic replay, workload-pressure replay reports, replay capacity envelopes, prefill/decode utilization, KV-page occupancy, vLLM/SGLang mirror normalization, streaming-token trace validation, route/scheduler provenance checks, canary/shadow/rollback gates, structured release-triage owner hints, model/backend numeric drift checks, fused Triton kernels, cache-controlled measurement, tracing/profiling direction, TTFT and decode-token p95 checks, KV memory-pressure reporting, model-version transition metadata, and tail-latency regression checks.
Accelerator Co-Design
ONNX compilation, explainable mixed-target partitioning, peripheral-aware energy sensitivity, SRAM/DRAM traffic analysis, RTL verification, FPGA synthesis, and low-precision datapaths.
Secure / Forward-Deployed AI
TS/SCI mission context, model authorization, audit evidence, policy controls, operational workflow design, observable delivery, and public-safe data discipline.
AI Reliability / Production
SLOs, Prometheus-compatible artifacts, incident runbooks, release gates, failure accounting, triage routing, deterministic fixtures, trace/replay tooling, token-path telemetry, token-budget limiter evidence, sanitized trace export, Grafana dashboard provisioning, resilience testing direction, and telemetry that makes regressions visible.
Systems / Quant Engineering
Deterministic matching, order-book mechanics, benchmarkable execution, Linux fundamentals, C++20, latency distributions, cross-language correctness gates, performance-counter direction, and cache-aware benchmarking direction.
Engineering Evidence
Signals backed by public work.
GPU Kernel Performance
Triton Kernel Lab validates fused RMSNorm and SwiGLU kernels against FP32 oracles, autotunes SwiGLU launch parameters, controls cache state, preserves raw timing samples, compares with PyTorch eager and torch.compile, and records cache-cold RTX 5070 Ti regression artifacts.
Hardware / Software Co-Design
HeteroCore links ONNX compilation, analog non-ideality simulation, SRAM/DRAM traffic modeling, synthesizable SystemVerilog, and FPGA schedule execution through a versioned plan while clearly separating projections, simulations, synthesis outputs, and future board measurements.
Inference Performance
Triton Inference Benchmark shows repeatable load generation, concurrency controls, p50/p95/p99 latency, throughput, retry and failure accounting, Prometheus output, baseline regression reports, correlated Triton/DCGM telemetry, exact-output batch-invariance CI gates, token throughput, requests per GPU-hour, and normalized cost-to-serve estimates.
Rust Inference Runtime
Rust Inference Runtime implements stable priority admission, bounded prefill work, conservative paged KV reservations, round-robin decode scheduling, deterministic trace fingerprints, and baseline/candidate promote, hold, and rollback decisions with queue-pressure, active-capacity, and replay capacity-envelope summaries, including 0.888889 decode-capacity utilization and 0.595062 KV-page occupancy in the pressure fixture, plus vLLM/SGLang-style mirrored observation normalization, streaming-token event validation, route/scheduler provenance coverage, model-aware numeric tolerance, segmented backend reports, TTFT and decode-token p95 checks, KV memory-pressure reporting, model-version transition metadata, token-trace fingerprints, structured triage for hold/rollback reasons, a checked streaming mirror promote report, synthetic fixtures, and CI.
Secure AI Service Boundaries
Secure GPU Inference Gateway shows authenticated model access, RBAC, reason-for-access policy, audit trails, request and token-budget limits, Prometheus metrics, estimated input-token throughput, sanitized trace JSONL evidence, synthetic capacity/cost planning tied to model policy, Grafana dashboard provisioning, SLO notes, incident runbooks, and production extension paths for policy-as-code, redaction, telemetry, and external authorization engines.
Mission Workflow Delivery
Readiness Control Tower turns synthetic sortie, maintenance, supply, personnel, and outage data into root-cause scoring, what-if analysis, recommendations, and command-facing workflows.
Deterministic Systems
Market Microstructure Engine demonstrates price-time priority, integer tick accounting, partial fills, cancellations, snapshots, a C++20 core, Python-oracle parity checks, and native latency measurement.
Experience
Cyber defense operations with measurable enterprise impact.
Cyber Defense Operations Specialist / Mission Assurance Technician / SolarWinds Support Technician
- Support ACC MAJCOM enterprise cyber operations in an 11-member team delivering 24/7 cyber command-and-control and tier-1 support to 26 wings across 13 CONUS bases, affecting 27K users.
- Engineered a cyber tracking tool that fused 216K data points into 4 live dashboards for a $1.3B enterprise environment, reducing threat resolution time from 5 days to 2.
- Led 11 Airmen across 18 AFCYBER tasking orders, hardening 5 vulnerable sites and 30K assets while supporting C4ISR readiness for 1.3K units across 262 locations.
- Formed a four-unit response team to detect, isolate, and contain a high-risk system anomaly, then helped shape updated cyber threat SOPs.
- Coordinated a three-agency inspection response, guiding 4 SMEs through 16 incident resolutions and helping secure the base's first successful cyber inspection rating in 9 years.
- Re-aligned a $1M modernization effort and synchronized 6 teams to field a network access control platform validating 16K users.
- Deployed SolarWinds Observability servers to ACC bases, identified rollout blockers, restored data paths, and strengthened enterprise situational awareness.
- Expedited 5 urgent change requests, shrinking execution from 21 days to 5 for a combat-readiness exercise supporting 10K warfighters and 409 tons of cargo.
- Authored a behavioral analytic for cyber risk and trained 14 squadrons to detect foreign-language anomalies across a $1.5B domain.
- Completed 180 hours of cyber and industry best-practice development, overhauled 3 SOPs, reduced system configuration time by 2 hours, and contributed to remediation of 486 exploits.
Developer / Systems Builder
- Build public-safe AI infrastructure and mission-software projects across Rust, Python, FastAPI, React, TypeScript, Docker, Kubernetes-oriented deployment patterns, metrics, tests, policy controls, benchmarks, telemetry artifacts, and runbooks.
- Build hardware/software co-design prototypes spanning ONNX graph partitioning, analytical cost models, analog non-ideality simulation, memory hierarchy analysis, SystemVerilog, self-checking RTL, synthesis, and FPGA schedule execution.
- Build deterministic market-systems work around order-book mechanics, price-time priority, a dependency-free C++20 core, Python-oracle parity checks, and p50/p95/p99 latency measurement.
- Administer Linux-hosted game server infrastructure using Java, Python, SQL, SSH operations, observability, permissions systems, and performance profiling for a 600+ member community.
Portfolio
Public-safe projects that show the engineering direction.
Readiness Control Tower
Synthetic mission readiness platform with FastAPI, React, root-cause scoring, what-if analysis, operational recommendations, Docker, tests, and command-facing workflows.
Secure GPU Inference Gateway
Security-focused AI infrastructure gateway with model authorization, reason-for-access checks, request and token-budget limits, audit logs, Prometheus metrics, estimated input-token throughput, sanitized trace export, checked trace and capacity-plan artifacts, Grafana dashboard provisioning, Kubernetes posture, SLO notes, incident runbooks, and a roadmap toward policy-as-code, redaction, and control evidence.
Triton Kernel Lab
Fused RMSNorm and autotuned SwiGLU kernels with FP32 correctness validation, explicit cache control, CUDA-event latency distributions, PyTorch compiler comparison, machine-readable reports, and a GPU regression gate.
Rust Inference Runtime
Deterministic continuous batching and paged KV-cache admission with stable priority ordering, round-robin decode progress, replay fingerprints, workload-pressure summaries, replay capacity envelopes, prefill/decode utilization, KV-page occupancy, vLLM/SGLang-style mirror normalization, streaming-token trace validation, route/scheduler provenance coverage, model-aware numeric tolerance, structured release triage, token-path telemetry, and canary/shadow promote, hold, and rollback gates.
HeteroCore Compiler
ONNX compiler and analytical cost model for mixed analog-digital inference, connected to analog simulation, memory hierarchy, synthesizable RTL, and FPGA schedule execution through a versioned plan.
Triton Inference Benchmark
Repeatable load-generation harness for Triton-style model serving, with concurrency controls, p50/p95/p99 latency, throughput, retries, Prometheus export, baseline regression gates, exact-output batch-invariance checks, token and GPU-capacity metrics, explicit cost assumptions, normalized cost-to-serve estimates, and Kubernetes job posture.
Market Microstructure Engine
Python and C++20 matching engines covering price-time priority, partial fills, market orders, cancellations, deterministic parity checks, native tests, and latency distributions.
Skills
Technical range across AI infrastructure, security, systems, and mission software.
Cyber / Security Ops
SolarWinds Observability, ServiceNow, Remedy, eMASS, CAD, ACAS, Tanium, ARAD, vulnerability management, incident response, compliance tracking, task order workflows, POA&M/MTO workflows.
Backend / Product
Rust, Python, TypeScript, Java, SQL, Bash, FastAPI, React, REST APIs, dashboards, automation scripting, data analysis, tests, and reviewer-friendly documentation.
Infrastructure / Observability
Linux, Docker, Git, Kubernetes deployment shapes, Prometheus-compatible metrics, OpenTelemetry direction, SLOs, release gates, token-budget limiting, token-path observability direction, resilience testing direction, runbooks, sanitized trace export, Grafana dashboard provisioning, rollback thinking, routing and switching, enclave support, SSH operations, and troubleshooting.
AI / Systems
AI infrastructure, inference benchmarking, model-serving concepts, latency percentiles, throughput, token and GPU-hour capacity, cost-to-serve modeling, failure accounting, routing and batching concepts, PyTorch, torch.compile, Triton GPU kernels, launch autotuning, cache-controlled CUDA-event measurement, CUDA/NCCL concepts, vLLM/SGLang-style mirrored observation adapters, streaming-token trace validation, route/scheduler provenance coverage, routing/scheduling/autoscaling concepts, workload-pressure replay summaries, replay capacity-envelope reports, prefill/decode utilization, KV-page occupancy, instrumentation/tracing/profiling direction, TTFT and decode-token p95 checks, KV memory-pressure reporting, model-version transition metadata, release-triage owner hints, accelerator-aware release gates, ONNX, SystemVerilog, FPGA flows, analytical accelerator and memory models, GPU-aware reliability, Rust scheduling, numeric release gates, segmented backend validation, C++20, cross-language oracle testing, Linux performance-counter direction, cache-aware benchmarking direction, and performance profiling.
Recognition
Awards, certifications, and education.
Awards
- HQ ACC/A6 Airman of the Year, 2025
- HQ ACC/A6 Team of the Year, 2025
- HQ ACC/A6 Warfighter Communication Award, 2025
- HQ ACC/A6 Airman of the 3rd Quarter, 2024
- 4x Squadron Team of the Quarter
- HQ ACC/A6 1-star coin
Certifications & Training
- CompTIA Security+
- CompTIA A+ ce Certification
- AWS Technical Essentials
- Introduction to Cloud Financial Management
- Generative AI Essentials on AWS
- Introduction to Generative AI on AWS
Education
- Western Governors University, B.S. Cybersecurity and Information Assurance, in progress
- Northern Virginia Community College, Computer Science coursework, 60+ credits completed
Leadership & Service
- Led Booster Club concessions committee, dedicating 80 hours and raising funds for morale events.
- Volunteered at a local food bank, packaging 11K pounds of food for the VA community.
- Supported AFCEA events by escorting distinguished visitors on base.
- Augmented a Base Appearance Team, helping clear roughly 1.5K pounds of storm debris.
Contact
Public profile first, sensitive details by request.
This public resume site intentionally keeps phone and direct email off-page. Recruiters and hiring teams can reach me through LinkedIn or review the portfolio through GitHub.