Selected project evidence

Projects explained through context, decisions, and outcomes.

Case studies spanning distributed media, analytics, digital agreements, omnichannel customer service, cloud cost optimization, engineering systems, applied AI, and this portfolio platform.

Flagship client platform

Multi-tenant omnichannel customer service platform

A client-owned platform that consolidates email, SMS, website chat, and phone workflows across multiple brands, with skill-based agent permissions and reviewable AI assistance.

Architecture, product and workflow design, full-stack implementation, database and AWS design, AI integration, team leadership, and executive communication
2026–present

Problem

A large U.S. pool-supply distributor operated multiple brands across several disconnected service systems with different rules, SLAs, and communication methods.

Constraints

Client identity remains confidential
Agents may support several brands without confusing data or queues
The initial architecture targets growth from roughly 1,000 to 10,000 tickets per month
AI-written customer responses must remain reviewable before sending

Architecture and decisions

Unified email, SMS, chat, and phone into one platform
Separated brand data and enforced backend permissions with skill-based agent access
Made brand context explicit through labels, colors, and separate queues
Added AI routing and classification with deterministic department fallbacks
Kept suggested replies, clarity edits, and translation behind human review
Isolated channel workloads while accepting PostgreSQL as a remaining shared dependency

Outcomes

Currently operates across five brands
Ticket handling and response times decreased
Manual work and recurring service costs decreased
The client owns the system rather than paying a recurring platform subscription

Distributed platform

StudioPlayer distributed content and device platform

A multi-tenant web and edge-device platform that lets organizations remotely schedule and manage media across distributed screens, including offline playback and event-driven casino installations.

Full-stack engineer, sole builder of the Linux/Python player, and current technical owner
2020–present

Problem

Multi-location organizations had to update televisions, LED boards, and outdoor signs manually—sometimes requiring travel or physical lifts.

Constraints

Devices operate outside the team’s direct physical control
Playback must continue through internet outages
Customer and casino identities and integration partners remain confidential

Architecture and decisions

Built a multi-tenant React and Node.js content-management system with schedules, permissions, and media lifecycle controls
Individually designed and built the Linux/Python mini-computer player
Used local media storage so the current playlist continues offline
Supported HTTP polling or WebSocket updates by player generation
Expanded the platform with closed-loop event servers coordinating media, audio, and lighting for casino events

Outcomes

Approximately 15 active customer organizations and more than 100 active devices
Thousands of media assets and daily device updates
One customer location can update more than 50 devices concurrently
Original customers remain on the platform approximately six years later
Outdoor and distributed displays can be updated remotely instead of through manual physical access

Data and analytics platform

Studioease Analytics

A real-time and historical analytics system that transforms Mindbody data into actionable retention, conversion, membership, attendance, marketing, and customer-value reporting.

Database infrastructure, ingestion, AWS, and backend API owner; current team lead
2022–present

Problem

Studio owners had to pull operational data manually and maintain spreadsheets without a custom analytics interface.

Constraints

Source data originates in Mindbody
Hundreds of millions of records require integrity checks and repair paths
Realtime ingestion currently shares the operational database with reads and other writes

Architecture and decisions

Built Mindbody API and webhook ingestion plus nightly Lambda reconciliation
Designed MySQL database infrastructure and Node.js/Express backend APIs
Added self-healing and re-sync workflows
Began designing a buffered ingestion pipeline to isolate high-volume writes; that pipeline remains in development

Outcomes

Processes hundreds of thousands of incoming records per day
Stores hundreds of millions of records for a few dozen customers
Produces live dashboards plus daily, weekly, and monthly reporting
Replaced recurring manual spreadsheet pulls with more consistent operational reporting

Secure workflow platform

Studioease E-Sign

A multi-tenant digital agreement workflow with versioned templates, expiring signer access, immutable signed records, audit metadata, and encrypted document storage.

Backend API, data model, audit trail, document storage, recovery, and PDF workflow builder
2022–present

Problem

Studios needed an integrated, lower-cost alternative to external signature tools while retaining control of their agreement data and workflows.

Constraints

Signed records must preserve the exact template revision
Customer links need bounded access
The public case study must not claim legal or regulatory compliance without a formal review

Architecture and decisions

Created UUID-versioned templates with signature, initial, guardian, text, and branding fields
Made completed signature records immutable
Captured browser/device, IP, date, location, and customer audit metadata
Stored final PDFs in encrypted S3 and delivered them through 24-hour presigned URLs
Protected employee access through the authenticated application and rate-limited expiring signer links

Outcomes

Signs and stores hundreds of documents and agreements daily
Reduced reliance on more expensive external signature services
Kept agreement data and downstream integrations inside the product workflow

Applied AI · Internal deployment

Agentic engineering platform

A deployed internal system that coordinates bounded agents for greenfield builds, bug fixes, feature work, and production-readiness review while keeping high-risk actions behind human approval.

Originator, architect, and primary operator
2026–present

Problem

Engineers can spend too much time approving repetitive tool requests and transferring prompts instead of reviewing the quality of the resulting work.

Constraints

Deployed internally with Tyler as the current primary operator
Broader team adoption and formal productivity telemetry remain ongoing
Destructive actions, AWS changes, production merges, and major architecture changes require human approval

Architecture and decisions

Created separate greenfield, bug-fix, feature, and deployment-readiness workflows
Used branch isolation, command restrictions, secret scanning, automated tests, human-in-the-loop hooks, and deployment gates
Persisted handoff documents and bounded memory so work can stop and resume

Outcomes

Initially deployed in June 2026 and used across more than ten real company repositories
Agent-assisted refactors have been reviewed, merged, and released to clients
Personal productivity gains are substantial, while formal team-wide telemetry remains ongoing

Engineering leadership

Engineering operating system

A practical transformation of code review, testing, deployment, security, documentation, AI usage, and work planning for a small engineering organization.

Executive owner and hands-on implementer
2025–present

Problem

The organization had little formal code review, testing, deployment control, security review, documentation, or consistent project planning.

Constraints

Small startup team
Engineers previously worked in project silos
Urgent client work still requires a controlled hotfix path

Architecture and decisions

Reduced project silos so peers can review one another’s work
Added GitHub approval protections and a defined hotfix exception
Expanded automated, regression, integration, end-to-end, security, and AI-oriented testing
Implemented New Relic database anomaly monitoring and more consistent security and database reviews
Created a formal AI policy and internal guidance
Introduced reviewed and estimated backlog work with priority tiers

Outcomes

Engineering work is more focused
More tickets reach full completion without being lost
Review, testing, and deployment expectations are explicit rather than improvised

Cloud operations and FinOps

AWS cost and infrastructure optimization

A production-account review that cut monthly AWS spend from a nearly $8,000 peak to $4,000 or less in each subsequent month while improving infrastructure discipline.

Review, analysis, implementation, and ongoing owner
2024–present; quantified intervention in 2026

Problem

AWS costs had accumulated gradually through oversized compute and databases, poor workload-to-database fit, and unused resources.

Constraints

Changes affected live production systems
Cost Explorer evidence is internal
Savings must not be annualized without a full year of measurements

Architecture and decisions

Reviewed the full account and connected systems
Rightsized EC2 and RDS resources
Matched database deployment models and capacity to actual workload behavior
Removed unused resources
Changed systems incrementally outside peak periods
Improved secrets, monitoring, deployment controls, and adoption of AWS CDK

Outcomes

Monthly AWS spend peaked near $8,000 in March 2026
Spend remained at or below $4,000 in every subsequent month through July 2026
Ongoing reviews now track usage and additional StudioPlayer optimization opportunities

Security and incident response

SQL injection incident response and remediation

Containment and remediation of an attempted SQL injection attack discovered through anomalous database behavior, followed by client reporting and stronger preventive controls.

Incident owner, technical lead, and client communicator
2025

Problem

A database performance review revealed unexpected slowness caused by probe traffic against exposed SQL endpoints.

Constraints

The issue affected a live client system
Containment required balancing availability against the risk of continued probing
Customer and company names remain confidential

Architecture and decisions

Temporarily stopped affected services to contain the attack path
Blocked the attacker's access and coordinated repair of the exposed endpoints
Verified the incident scope and confirmed that no customer data was accessed
Prepared the complete incident report for the client
Led a blameless but accountable engineering review focused on prevention

Outcomes

The attempted attack was contained without customer data access
The vulnerability was repaired before service restoration
The incident informed stronger database monitoring, security review, and team guidance

Platform architecture · Applied AI

Evidence-backed portfolio and career intelligence platform

Designed and built a secure, serverless portfolio platform that connects verified career content, recruiter experiences, and visibility-aware AI retrieval.

Product owner, solution architect, and hands-on engineer
2026–present

Problem

Create a professional portfolio that operates as an evidence-backed career intelligence platform rather than a static résumé.
Make career evidence useful to technical recruiters, hiring managers, engineering leaders, and software engineers without changing the underlying facts for each audience.

Constraints

Keep expected AWS operating costs within a $50 monthly ceiling.
Protect administrative access, unpublished content, private documents, and AI features from unauthorized use.
Keep the public portfolio useful when AI features are unavailable.
Support secure, accessible, and mobile-friendly experiences.

Architecture and decisions

Separated public, authenticated administration, and asynchronous worker boundaries, with server-side authorization on every administrative API.
Selected usage-based AWS services including CloudFront, Cognito, Lambda, API Gateway, DynamoDB, SQS, Amazon Bedrock, and S3 Vectors.
Implemented visibility-aware RAG ingestion through an encrypted FIFO queue with bounded concurrency, retries, a dead-letter queue, and an explicit AI cost switch.
Defined infrastructure with AWS CDK and deployed through GitHub Actions using short-lived OIDC credentials instead of stored AWS access keys.
Used a durable architecture document, repository guidance, and project-specific agent skills to keep agent-assisted development consistent across sessions.

Outcomes

Deployed a private HTTPS staging environment with authenticated content administration and responsive public and administrative interfaces.
Established CMS-managed publication, placeholder, evidence, page-visibility, and chat-visibility controls.
Activated the staging RAG pipeline and verified that all 40 unapproved placeholder records were excluded without generating embeddings or Bedrock usage.
Established reproducible infrastructure releases, least-privilege workload boundaries, finite logs, rate and concurrency controls, a $50 AWS budget, and a public AI kill switch.

Continue the evaluation

Have a similar system or team challenge?

Use Ask Tyler to explore the relevant evidence or analyze your role against this portfolio.

Ask Tyler Analyze a role