AI Engineering Portfolio

AI engineering for structured, reviewable business workflows.

I design AI systems that turn operational inputs into validated outputs, traceable decisions, review states, and business-ready workflows.

Structured outputs
Validation and review gates
Workflow data quality
View Case StudiesTechnical ApproachContact
José Velásquez with AI-assisted real estate analysis interface
01

Operating Context

Built from real workflow exposure.

My work sits at the intersection of commercial real estate operations, structured data workflows, and governed AI systems.

Workflow Understanding

Brokerage operations, CRM updates, availability tracking, research support, data quality, and AI extraction UAT.

AI Engineering Discipline

Structured outputs, validation logic, review gates, workflow state, persistence, and audit-ready delivery.

Business Adoption

Systems that reduce cleanup work, expose evidence, handle exceptions, and give operators confidence.

02

What I Build

AI is useful when it is bounded, validated, reviewable, and measurable.

The strongest systems are not open-ended prompts. They are workflows with clear inputs, tool boundaries, evidence, review states, and operating metrics.

Data Quality Agents

Convert incoming business information into validated candidate updates with confidence and review states.

Source-Grounded Research

Turn selected evidence into structured, traceable outputs that can be reviewed and reused.

Decision Support Workflows

Combine scoring, rationale, benchmarks, and summaries into business-ready decision artifacts.

Agentic Workflow Architecture

Route tasks through tools, validation, persistence, audit logs, and human approval points.

03

Case Evidence

Two proof points: decision intelligence and research traceability.

Convertis and Ingeniometrix show the same engineering pattern in different domains: structured intake, bounded AI use, validation, persistence, and reviewable outputs.

Flagship Case Study

Archangel – Convertis

Convertis turns property inputs into structured real estate intelligence: validated intake, location enrichment, deterministic scoring, AI-supported interpretation, and live reviewable outputs.

Explainable scoring
Validated outputs
Human review
CRE IntelligenceExplainable ScoringStructured OutputsLive Workflow
Open Case StudyLaunch Live Project

Research Workflow

Ingeniometrix Research Assistant

Ingeniometrix combines structured intake, academic source retrieval, selected evidence, strict JSON schema generation, coherence validation, and audit-ready outputs.

Source-grounded generation
Traceability validation
Review-ready output
Source GroundingStrict JSON SchemaTraceabilityWorkflow State
Open Case StudyLaunch Live Project
04

Agent Concept

Brokerage Data Quality & Intake Agent.

A practical AI agent pattern for converting unstructured brokerage information into validated, reviewable CRM candidate updates.

AI proposes. The workflow validates. People approve.

The concept focuses on reducing data cleanup work while preserving evidence, field-level confidence, exceptions, audit logs, and measurable value.

Compare AI Patterns

Inputs

  • Listing flyers
  • Broker emails
  • CRM events
  • Property, suite, broker, and contact records

Tools

  • Document parser
  • CRM/database lookup
  • Broker/contact lookup
  • Validation functions

Outputs

  • Candidate CRM updates
  • Data-quality issue tickets
  • Broker-ready summaries
  • Review tasks

Metrics

  • Field accuracy
  • Human correction rate
  • Duplicate reduction
  • Update turnaround time
05

Technical Approach

My AI engineering approach.

I design AI workflows around trust: clear data contracts, visible evidence, deterministic checks, human review, and measurable operating value.

01

Start with workflow pain

Identify where data breaks, users lose time, and cleanup work accumulates.

02

Bound the input

Define source types, required fields, confidence signals, and validation rules.

03

Separate AI from control logic

Keep scoring, routing, validation, persistence, and review outside the model.

04

Expose evidence and confidence

Preserve source IDs, field-level confidence, rationale, and uncertainty.

05

Design review gates

AI proposes; people approve, reject, correct, or route exceptions.

06

Measure business value

Track accuracy, correction rate, turnaround time, latency, cost, and duplicate reduction.

06

Portfolio Map

Different domains, one governed AI pattern.

Each project emphasizes a concrete business workflow, a bounded AI pattern, a validation mechanism, and a reviewable output.

ConvertisCRE decision supportAI-assisted interpretationDeterministic scoringReviewable scorecard
IngeniometrixResearch planningSource-grounded generationTraceability and coherence checksBlueprint artifact
Data Quality AgentCRM and operationsAgentic extraction and validationField confidence and duplicate checksCandidate CRM update
Listing ExtractionCRE data extractionStructured extractionSchema and confidence flagsNormalized listing record
07

Additional Work

Supporting systems and concept work.

Additional case studies show operational systems thinking, extraction design, and cloud-agent architecture mapping.

Engineering QA

Simetrika Systems Experience

Led structural design, BIM coordination, technical validation, and structured documentation for large mid- and high-rise developments.

CRE Data Quality

AI-Assisted Listing Extraction

Designed and tested workflows for extracting structured listing data from flyers, including addresses, suites, rents, availability, brokerages, and contacts.

Agentic Automation

OpenClaw Agentic Assistant

Implemented OpenClaw in a Linux environment to explore persistent agentic workflows, remote monitoring, orchestration, and multi-step automation.

GCP Concepts

Cloud & Enterprise Mapping

Mapped applied app experience to enterprise AI patterns across Vertex AI concepts, Cloud Run, Cloud SQL, monitoring, latency, and cost tradeoffs.

08

About

Engineering discipline applied to AI systems.

I started in civil and structural engineering, where complex systems, modelling, validation, and cross-functional execution were central to the work.

Systems mindset, business context, AI execution.

I now apply that same systems mindset to AI engineering, with a focus on structured workflows, operational data quality, commercial real estate processes, and reviewable AI outputs.

Designed for people who need to trust the output.

My current work gives me practical exposure to brokerage operations, research support, CRM data maintenance, AI extraction UAT, and data-quality workflows. That context shapes how I design AI systems: useful, bounded, measurable, and reviewable.

09

Technical Focus

Capabilities behind the case studies.

The homepage stays business-oriented; the case studies provide the deeper technical evidence, architecture, and code excerpts.

AI Systems

  • LLM workflows
  • Structured outputs
  • Source-grounded generation
  • Agentic tool use
  • Prompt and version discipline

Validation & Trust

  • Schema validation
  • Traceability checks
  • Field-level confidence
  • Review states
  • Audit logs

Data & Backend

  • Python
  • TypeScript
  • SQL
  • PostgreSQL
  • APIs
  • Prisma and Supabase concepts

Cloud & Enterprise Concepts

  • GCP
  • Vertex AI concepts
  • Cloud Run
  • Pub/Sub and Cloud Tasks
  • Cloud SQL
  • Logging and monitoring
View the case evidence