Architecture portfolio

Things I've Built

Systems where I was responsible for the technology architecture, with hands-on involvement across product engineering, deployment, DevOps, cost, scale, reliability, and production readiness.

Client names are intentionally omitted where confidentiality matters. The architecture, outcomes, and lessons are preserved without exposing private customer identities.

Architecture Responsibility

System Design

Architecture choices, service boundaries, data flows, integration patterns, and hands-on technical direction.

Deployment & DevOps

CI/CD, infrastructure automation, cloud deployment, release readiness, and operational controls.

Cost & Scale

Runtime efficiency, capacity planning, cost optimization, and scaling patterns aligned to workload needs.

Reliability

Observability, high availability, security posture, incident readiness, and production resilience.

Founder-Built Products

3 builds captured in this phase

AIOps Platform · 2025-Present

OpsRabbit

AI-powered CloudOps and SRE incident investigation platform.

AIOpsSRECloudOpsIncident AutomationReliability Engineering

Role and Outcome

Responsible for technical architecture and hands-on delivery direction across system design, deployment, DevOps, cost, scale, reliability, and production readiness.

Built OpsRabbit as a hands-on founder product focused on reducing MTTR, improving incident investigation, and helping engineering teams connect alerts, logs, deploys, ownership, and operational context.

View detailed build note

Scale

Built as a founder-led product initiative for modern SRE, CloudOps, and incident response teams.

Architecture

Designed the product around AI-assisted incident investigation and operational evidence gathering.
Focused workflows on CloudOps, SRE, incident response, and reliability engineering use cases.
Built the platform direction around practical automation that helps teams move from noisy alerts to clearer action.
Kept the product grounded in real operational workflows rather than generic chatbot-style responses.

Lessons Learned

AI for operations must connect incidents to real engineering context: ownership, deploys, alerts, logs, and system behavior.
SRE automation needs explainability and traceability because engineers will not trust black-box answers during production incidents.
The product architecture has to fit existing incident workflows while gradually improving speed, consistency, and reliability.

AI SaaS Product · 2019-2024

7Targets AI Sales Assistants

AI-assisted sales automation platform for follow-ups, response handling, and sales workflow productivity.

AI SalesSaaSSales AutomationAI AssistantsFounder-Built

Role and Outcome

Responsible for technical architecture and hands-on delivery direction across system design, deployment, DevOps, cost, scale, reliability, and production readiness.

Hands-on built and scaled 7Targets as a SaaS product for AI-assisted sales follow-up, inbound/outbound response handling, and workflow automation for sales teams.

View detailed build note

Scale

Built as a multi-year SaaS product with founder-led product, engineering, and customer feedback cycles.

Architecture

Built AI assistant workflows for sales follow-ups, lead engagement, response handling, and sales activity automation.
Designed the product around practical sales-user workflows, customer feedback loops, and SaaS operating needs.
Worked across product, platform, and customer-facing implementation concerns as a hands-on founder and CTO.
Evolved the product with recurring feedback from smaller companies and sales teams using AI assistance in daily workflows.

Lessons Learned

Practical AI products win when the intelligence is embedded inside the user's daily workflow rather than exposed as a separate tool.
Founder-led SaaS architecture needs tight loops between customer feedback, product direction, engineering tradeoffs, and support reality.
Automation should improve consistency and follow-through while preserving the human intent behind communication.

Document AI Product · 2021-Present

DocuLens

AI document intelligence product for extraction, summarization, OCR, translation, redaction, and chat-over-documents workflows.

Document AIOCRData ExtractionRedactionAI Product

Role and Outcome

Responsible for technical architecture and hands-on delivery direction across system design, deployment, DevOps, cost, scale, reliability, and production readiness.

Hands-on built DocuLens as a document intelligence product to help users work with unstructured documents through OCR, extraction, summarization, translation, redaction, and document chat capabilities.

View detailed build note

Scale

Built as a hands-on AI product initiative for document-heavy business workflows.

Architecture

Built document-processing workflows around OCR, extraction, summarization, translation, redaction, and conversational document interaction.
Designed the product around practical document-heavy business workflows where users need faster understanding and safer handling of information.
Focused on turning unstructured document inputs into usable structured outputs and summaries.
Kept redaction and controlled document handling as important product concerns because document workflows often include sensitive information.

Lessons Learned

Document AI quality depends on the full pipeline: ingestion, OCR, extraction, summarization, redaction, and user-facing interaction.
Security and controlled handling need to be designed into document workflows early because business documents often contain sensitive data.
The strongest product experience comes from combining extraction, summarization, and interaction into one workflow instead of separate utilities.

Confidential Client Builds

4 builds captured in this phase

Agentic AI

B2B Customer Service Voice Bot 2.0

Customer service modernization for a B2B apparel and distribution business unit.

LangGraphLangChainVertex AIGKERAGVoice Bot

Role and Outcome

Responsible for technical architecture and hands-on delivery direction across system design, deployment, DevOps, cost, scale, reliability, and production readiness.

Modernized B2B customer service by automating complex inquiries across inventory, pricing, order tracking, and shipping. The system targeted 80% automation, reduced live agent load, and supported English and Spanish interactions.

View detailed build note

Scale

Designed for high availability and handled approximately 1,000 calls per month.

Architecture

Moved from a rigid Dialogflow-only implementation to a multi-agent architecture using LangGraph and LangChain.
Hosted the agentic system on a multi-node Google Kubernetes Engine cluster with multi-zone redundancy.
Used Vertex AI Gemini for generative responses and Vector Search for RAG-based FAQ retrieval.
Connected securely to legacy on-premise ERP systems through VPN/private tunnels for real-time lookups.
Implemented a dedicated PII redaction layer before persistence or logging.

Lessons Learned

Intent-only bot architectures become brittle as business workflows grow; agentic systems need modular boundaries that can be observed and debugged.
Production LLM systems require traces, metrics, and prompt/output visibility from day one, otherwise reliability and cost cannot be managed.
For enterprise AI, architecture must define safe limits clearly: read-only actions, escalation paths, and human control for sensitive workflows.

Cloud AI Platform

AI Sandbox Azure Deployment

Secure AI experimentation environment on Microsoft Azure.

AzureAI SandboxTerraformContainersCloud Security

Role and Outcome

Responsible for technical architecture and hands-on delivery direction across system design, deployment, DevOps, cost, scale, reliability, and production readiness.

Delivered a functional AI Sandbox on Azure so the business unit could test and validate AI models without affecting production systems.

View detailed build note

Scale

Built as a dedicated business-unit environment for safe AI model experimentation and validation.

Architecture

Used Azure-native cloud services for containerized deployment.
Created an isolated sandbox that mirrored production-grade security constraints.
Kept the design portable across clouds by standardizing infrastructure-as-code practices.

Lessons Learned

Multi-cloud architecture is easier to sustain when infrastructure patterns are standardized through Terraform and repeatable delivery practices.
A sandbox is useful only when it reflects production security, networking, and deployment constraints closely enough to reveal real risks.
Cloud experimentation environments still need operational discipline around access, cost, isolation, and lifecycle management.

Agentic AI

Marketing Offer Intelligence Agent

AI agent for marketing teams to analyze public market offers and create standardized offers faster.

Agentic AIMarketing AutomationOffer IntelligencePublic DataTemplate Standardization

Role and Outcome

Responsible for technical architecture and hands-on delivery direction across system design, deployment, DevOps, cost, scale, reliability, and production readiness.

Built an AI agent that helped marketing teams review publicly available offers from banks and financial institutions, compare market patterns, standardize offer templates, and create new offers more quickly.

View detailed build note

Scale

Built for marketing research and offer-generation workflows where teams needed faster comparison across public market data.

Architecture

Designed standardized offer templates so marketing teams could move from ad hoc offer creation to repeatable structured workflows.
Collected and organized publicly available offer data from banks, financial institutions, and market sources.
Used AI-assisted summarization and comparison to highlight offer patterns, positioning, eligibility, benefits, and terms.
Created a workflow that helped users move from market research to draft offer creation with less manual effort.

Lessons Learned

AI agents create business value when unstructured market research is converted into a repeatable operating workflow.
Public-data workflows need structure, traceability, and review paths so users can understand how recommendations were derived.
The architecture should standardize the business artifact first; data collection and AI generation become much more reliable after the template is clear.

DevSecOps

GitLab DevSecOps Modernization

Security-first modernization of CI/CD, infrastructure, and service delivery practices.

GitLabDevSecOpsGoogle Secret ManagerWorkload IdentityISO 27001NIST

Role and Outcome

Responsible for technical architecture and hands-on delivery direction across system design, deployment, DevOps, cost, scale, reliability, and production readiness.

Moved the organization toward a security-first CI/CD model, reducing manual overhead and strengthening alignment with ISO 27001 and NIST-oriented controls.

View detailed build note

Scale

Designed as a foundational delivery platform for multiple application and infrastructure workflows.

Architecture

Centralized source control and CI/CD on GitLab.
Enforced signed commits, stronger secret handling, and secure pipeline controls.
Used Google Secret Manager for secret management.
Enforced TLS 1.2+ across service endpoints.
Embedded static and dynamic security checks inside the deployment pipeline.

Lessons Learned

Security becomes sustainable when it is built into CI/CD pipelines, identity, secrets, and infrastructure automation instead of reviewed after deployment.
Workload identity and short-lived access patterns reduce operational risk more effectively than static keys.
DevSecOps architecture should make secure delivery the default path for engineering teams, not a separate approval burden.

Growth Phase: Modernization and Scale

3 builds captured in this phase

Integration Platform · 2021-2022

Document Workflow Integration Gateway

Cloud-native integration gateway for ERP adapter configurations.

AWS LambdaStep FunctionsDynamoDBReactERP Integration

Role and Outcome

Responsible for technical architecture and hands-on delivery direction across system design, deployment, DevOps, cost, scale, reliability, and production readiness.

Created a generic API integration gateway that reduced maintenance by using a single cloud-native codebase across multiple ERP adapter configurations.

View detailed build note

Scale

Built to support multiple ERP adapter configurations from a shared integration codebase.

Architecture

Used an AWS serverless-first architecture with Lambda and Step Functions for orchestration.
Stored integration state and configuration data in DynamoDB.
Built a React frontend to manage integration workflows and adapter configuration.
Designed the platform around reusable adapters instead of bespoke client-specific integration code.

Lessons Learned

Integration architecture should avoid bespoke code where configuration-driven adapters can provide reuse and lower maintenance.
Serverless workflows are powerful when the integration stages, retries, state, and operational ownership are explicit.
A generic adapter platform reduces delivery cost only when it is paired with clear templates, testing, and support boundaries.

Cloud Modernization · 2022-2024

Recruiting Platform Infrastructure Modernization

Legacy infrastructure modernization, security improvement, and AWS organization consolidation.

AWS ECS FargateSpot InstancesTerraformAWS SSOCost Optimization

Role and Outcome

Responsible for technical architecture and hands-on delivery direction across system design, deployment, DevOps, cost, scale, reliability, and production readiness.

Transformed legacy infrastructure into a secure, cost-optimized platform, including AWS organizational consolidation.

View detailed build note

Scale

Modernized a production platform over a multi-year growth period while improving cost, security, and operational consistency.

Architecture

Used ECS Fargate with Spot capacity to improve cost efficiency.
Standardized infrastructure management with Terraform.
Implemented G-Suite and AWS SSO integration for identity access.
Consolidated AWS organization structure to improve governance and operational control.

Lessons Learned

Cost optimization works best when it is designed into runtime architecture, capacity choices, and deployment automation from the start.
Terraform changes cloud operations from manual activity into governed, reviewable, repeatable engineering.
Identity, account structure, and access governance are core architecture concerns in any serious modernization program.

Applied AI · 2022

AI/ML Video Confidence Predictor

Computer-vision workflow for confidence scoring from video signals.

AWS SageMakerMediaPipeMoviePyLambdaStep FunctionsComputer Vision

Role and Outcome

Responsible for technical architecture and hands-on delivery direction across system design, deployment, DevOps, cost, scale, reliability, and production readiness.

Automated facial and posture expression analysis to derive High, Medium, and Low confidence scores.

View detailed build note

Scale

Built as an automated video-analysis pipeline for repeatable confidence scoring.

Architecture

Used AWS SageMaker for model experimentation and ML workflow support.
Used MediaPipe and MoviePy for video processing and feature extraction.
Automated processing with AWS Lambda and Step Functions.
Designed the flow to convert raw video inputs into structured confidence classifications.

Lessons Learned

Applied AI systems need deterministic preprocessing and clear scoring logic so model outputs remain explainable and operationally useful.
Serverless orchestration works well for staged AI/ML workflows when each processing step has clear inputs, outputs, and failure handling.
AI architecture should make the business signal visible, not just run a model and return a score.

Recent Phase: Agentic AI and Intelligence

2 builds captured in this phase

Cloud Platform · 2024

IoT Hospitality Platform High Availability

High-availability platform work for a multi-tenant IoT hospitality system.

AWS EKSHelmTraefikCertManagerIoTHigh Availability

Role and Outcome

Responsible for technical architecture and hands-on delivery direction across system design, deployment, DevOps, cost, scale, reliability, and production readiness.

Enabled high availability for a multi-tenant IoT hospitality platform.

View detailed build note

Scale

Built for a multi-tenant IoT hospitality platform with high-availability requirements.

Architecture

Used AWS EKS as the Kubernetes runtime for service orchestration.
Managed complex service deployment and configuration through Helm charts.
Used Traefik and CertManager for secure, automated service communication.
Focused on tenant-aware operations, service reliability, and certificate lifecycle automation.

Lessons Learned

Kubernetes platforms become reliable when deployment, routing, certificates, observability, and service configuration are managed as one operating model.
Helm discipline matters because unmanaged configuration complexity becomes release risk in multi-service environments.
High availability is not only infrastructure redundancy; it also requires release hygiene, certificate automation, and operational ownership.

Agentic AI · 2025

Gen-AI and Agentic Framework QA Standards

Production-grade QA standards for LLM and RAG pipelines.

RAGASLangfuseCI/CDRAG EvaluationLLM QAAgentic AI

Role and Outcome

Responsible for technical architecture and hands-on delivery direction across system design, deployment, DevOps, cost, scale, reliability, and production readiness.

Established automated QA gatekeeping for LLM pipelines, moving beyond basic unit tests toward measurable recall, faithfulness, and relevance standards.

View detailed build note

Scale

Designed as a production QA standard for Gen-AI and agentic delivery pipelines.

Architecture

Integrated RAGAS into CI/CD pipelines to score retrieval recall, faithfulness, and answer relevance.
Configured hard build gates around quality thresholds such as 0.90 recall and 0.85 faithfulness/relevance targets.
Used Langfuse for live tracing, latency monitoring, and token cost tracking in production.
Connected evaluation signals to release decisions so quality regressions could block merges.

Lessons Learned

Gen-AI systems need evaluation architecture, not only unit tests, because quality depends on retrieval, prompts, grounding, and model behavior.
RAG pipelines should have measurable CI gates for recall, faithfulness, and relevance so quality regressions block release.
Tracing, evaluation, cost monitoring, and release control must work together before LLM systems can be treated as production-grade.

Selected Architecture Notes

1 build captured in this phase

Internal AI Operations

AI Scrum Master Elisa Rollout

Internal process improvement integrated into project delivery workflows.

AI AgentsScrumJiraSlackDelivery Automation

Role and Outcome

Responsible for technical architecture and hands-on delivery direction across system design, deployment, DevOps, cost, scale, reliability, and production readiness.

Automated routine scrum support, backlog flow, sprint reporting, meeting progress insights, pre-reads, and summary reports.

View detailed build note

Scale

Rolled into active project delivery to improve recurring team operations and visibility.

Architecture

Integrated AI-assisted workflow support into team ceremonies and project reporting.
Connected operational context from tools such as Jira and Slack.
Used centralized configuration management to reduce breakage from process and meeting changes.

Lessons Learned

Internal AI agents need the same product discipline as external tools: onboarding, data quality, reliability, and user trust determine adoption.
Workflow automation must handle configuration drift because real teams constantly change links, ceremonies, tools, and processes.
AI delivery automation is most useful when it improves team rhythm and decision-making, not just when it produces reports.