Enterprise technology consulting background — cloud migration strategy, post-M&A assessments, application rationalization, and TCO modeling for Fortune 500 clients across healthcare, financial services, retail, CPG, and government. Work spans a $651K UCaaS DC consolidation and a HIPAA-compliant multi-agent Bedrock platform to a production ML weather system and a three-cloud AI ecosystem at under $100/month.
Healthcare, financial services, and technology clients operating under compliance constraints don't have the option of deploying AI that can't be audited. These projects were scoped with governance as a design input — not a retrofit. Multi-agent architectures, HIPAA-compliant infrastructure generated from spec, dual-cloud deployment with zero forked business logic, and AI governance platforms that enforce SDLC gate reviews across 10 regulatory frameworks.
Multi-agent AI pipeline that scores an enterprise application portfolio against 6R cloud migration criteria and produces auditable, provenance-tagged recommendations — deployed for a healthcare imaging organization (36 clinical apps, 408 VMs, $24 total AI compute, 22 governance rules, zero black-box outputs). The full deployment stack behind it — CloudFormation on AWS, Bicep on Azure, Bedrock and GPT-4o multi-agent swarms, MCP servers, PHI-masking ETL — was generated from a written spec using Kiro with zero hand-written IaC, and runs identically cross-cloud with only a cloud-adapter layer separating the two versions.
Enterprise AI governance platform that scans codebases against 10 regulatory frameworks (including ISO 42001, NIST AI RMF, HIPAA), generates compliance artifacts, and enforces human approval gates at 5 SDLC checkpoints. The v3 release adds a Framework Explorer, a Policy-as-Code engine, and a full runbook library — DevSecOps, PM/Scrum, Workshop Guide, and Software Factory — so AI-generated commits are scanned against the same rules as human-written code. Deployed cross-cloud with zero forked business logic — AWS (CloudFormation, DynamoDB, Bedrock) and Azure (Bicep, Cosmos DB, Managed Identity) run identical application behavior behind a cloud-adapter layer — and productized as a multi-tenant SaaS (PostgreSQL + Redis, blue-green deploys, hybrid SDK) for teams that want the library without owning the infrastructure.
Pre-sales research agent that turns a company name into a technology diligence brief in under 10 minutes — live news, website analysis, stock data, stakeholder mapping, tech stack profiling, and opportunity framing. Replaces 4+ hours of manual pre-call research per account. Built for enterprise technology sales cycles where showing up informed is a competitive differentiator.
Collects 30 days of live datacenter telemetry via PowerShell/Bash agents, builds an in-memory dependency graph, and runs 5 AI simulations — wave rehearsal, blast radius, co-dependency discovery, right-sizing, and TCO projection — across 1,053 servers in 4 DCs. Used in enterprise cloud migration advisory engagements to give infrastructure teams a visual model of what a migration actually disrupts before they start.
Location-aware generative AI product that turns any GPS coordinate into a narrated historical brief, drawing on 60M+ digitized public-domain records from the Library of Congress, National Archives, Smithsonian, and DPLA. Serves 5 distinct audiences — consumers, K-12 educators, homeschoolers, tourism boards, and cultural institutions — from a single platform, and is positioned to qualify for NEH/IMLS cultural-heritage digital-access grant funding.
New enterprise data AI pipeline that connects to just about any enterprise data source — chat, ticketing, wikis, document repositories, email, or a custom pipeline — through a library of connectors, then runs every document through a 4-stage Bronze → Silver → Gold enhancement pipeline: deterministic pruning, Bedrock Guardrails screening, tool-forced Claude classification (temp=0), and embedding into a hybrid search index. Bedrock Guardrails and Claude are already the production AI layer today, called from Azure-hosted infrastructure — this documents the full AWS-native port target.
Every architectural pattern I recommend to enterprise clients has already run in production here first. Park Whisperer is a fully operational AI ecosystem — real users, live data pipelines, a six-model ML stack, automated content operations, and a conversational agent — running across Azure, GCP, and AWS at under $100/month. The domain is theme park intelligence. The purpose is proving that specific architecture decisions work before scaling them into environments where failure is expensive.
Full platform architecture →Production GCP pipeline that fuses METAR surface observations, NWS forecasts, and upper-air radiosonde soundings in BigQuery. ML scoring maps current conditions to per-attraction OPERATE / MONITOR / CLOSE recommendations and publishes a live JSON API to GCS every 15 minutes, consumed by the agent and dashboards.
ETL pipeline and SLM-routed query engine over 1.6M+ theme park operational records. A fine-tuned Phi-3 Mini classifier routes queries to one of 3 strategies (vector, numeric, hybrid). Deterministic fast paths bypass AI inference entirely for wait-time and outage questions — no latency, no hallucination.
Scheduled pipeline that produces and publishes theme park social content with no human involvement. The architecture: intent classification → parallel knowledge pre-fetch → LangChain agentic loop with 30+ tools → 3-pass generation → direct platform publish. Ten content types, zero hallucinated facts, grounded entirely in live operational data. The production system that validates the content generation patterns recommended in enterprise AI deployments.
Four AI models with distinct roles collaborate on each content piece. Phi-3 Mini classifies intent. Haiku runs the agentic tool loop (fast, cheap). Sonnet writes the final output — replacing Haiku after model-specific testing revealed it lacked domain knowledge for the subject matter. Each model has a persona and guardrails.
Five systems take a Claude-written script from text to a published social post. ElevenLabs per-character timestamps drive frame-accurate word-pop captions. Pillow composes brand overlays. FFmpeg encodes. An API publisher posts to three platforms. The pipeline was built to validate automated content production patterns — the same approach applies in enterprise content operations and AI-assisted marketing workflows at scale.
Conversational agent that answers questions about park operations in plain language. Uses the same Azure Function, SLM classifier, and 15+ live data tools as the content pipelines — the agent and content factory share infrastructure. Designed for responses short enough to paste directly into social DMs and comments.
Containerized ETL job running every 10 minutes that collects live attraction status, wait times, Lightning Lane sellout events, and schedules for 286 WDW entities and writes them to PostgreSQL. Feeds every downstream consumer: the RAG, the agent, and all content pipelines. The foundation of the platform.
How the platform's analytics store went PostgreSQL → Cosmos DB → back to PostgreSQL. A Cosmos DB monolith developed a critical hot-partition problem at 5.9M documents. The fix cost $394/month with no benefit. This documents the decision process, the migration, and the six architectural lessons — including when not to use a document store.
8 years of Central Florida METAR surface obs fused with twice-daily radiosonde soundings (CAPE, Lifted Index, K-Index) trains 6 scikit-learn models that nowcast thunderstorm, precipitation, fog, and venue-impact risk every 20 minutes. A spatial boost via BigQuery ST_DISTANCE amplifies probabilities when confirmed active storms are within 15 miles.
Ingests GOES-18 Geostationary Lightning Mapper data every 20 seconds, applies a two-stage spatial filter to attribute flashes to specific parks, and drives a 5-state storm machine (CLEAR → SEVERE) that fuses satellite data, 17-station METAR surface obs, NWS alerts, and SPC categorical outlook to compute per-ride closure risk.
Each of these directly preceded a production component. They're documented because architecture decisions are more credible when you can show what was tried first and why it was replaced. The forecasting iterations show how compounding error and inference-time feature availability constrained the final design. The agent framework explorations show the tradeoff analysis that led away from Assistants API and n8n toward stateless Azure Functions.
Global LightGBM model across all WDW attractions using a recursive multi-step loop — predicted values fed back as lag features for subsequent horizons. Seven forecast intervals from 15 minutes to 3 hours. One of 8+ iterations in the wait-time forecasting R&D track, each informing the next architectural decision.
Per-attraction time-series forecasting with one Prophet model per ride, 50+ external regressors, and a 24-hour horizon written back to BigQuery. Reached v8 before being shelved — rolling window features unavailable at inference time degraded accuracy significantly.
Three-cloud exploration of Bedrock Agents and custom action group patterns. Wired three weather tools to a GCP Cloud Run backend via two integration architectures: direct Lambda action groups and Azure Logic Apps as cross-cloud backends. The tool design patterns and OpenAPI schema approach carried forward directly into the production agent.
Earlier agent architecture using the Azure OpenAI Assistants API (threads/runs model). Persistent assistant object, 16 function tools, server-managed thread state, Cosmos DB vector search over 9 knowledge types. Directly preceded the production Park Agent Chat.
"Walt" — a central host agent — coordinated seven specialized sub-agents, each named for a Disney character and owning a park knowledge domain. Two custom FastAPI microservices ran alongside n8n. Iterated through 7 Docker Compose versions before being replaced by the Azure Functions architecture.
I take on a limited number of freelance engagements through AI Pathfinder LLC — architecture reviews, governed AI builds, cloud migration analysis, and hands-on delivery. If something in this portfolio is close to your problem, let's talk about yours.