From board strategy
to production code.
I'm Daniel Baez, CTO at Insurance Nation. I set technology strategy for a regulated insurance operation — and I architect and build the systems that deliver it: multi-cloud platforms, workflow automation, and AI document pipelines running in production.
Not looking for a role — always open to a conversation about modernization, governed automation, and applied AI in regulated operations.
About
I'm the CTO of Insurance Nation, where I lead the architecture and delivery of the platforms that run end-to-end insurance operations. The job spans both ends of the stack: I set technology strategy with ownership and the executive team, and I stay close enough to the code to make the architecture calls myself.
Originally trained as a Telecommunications Engineer in Havana, Cuba, I built my career across the full technology stack — from systems design into digital transformation leadership. I hold an MS in Computer Science from Georgia Tech and I'm an AWS ML Engineer – Associate.
Today the work is applied AI in a regulated industry: document intelligence, extraction from unstructured carrier paperwork, and workflow orchestration — built on the Claude and OpenAI APIs, AWS Bedrock and SageMaker, and locally hosted models for the cases where the data can't leave the building. In insurance, a model that is right most of the time isn't finished — so most of my attention goes to the parts that decide when not to automate: confidence thresholds, escalation paths, and which data is allowed to leave the building at all.
Outside of work I run pirabits.com, my personal experimental lab — a space for exploring AI, emerging technologies, and creative ideas outside the constraints of production systems. It's where I build and test without every idea needing to become a product.
Featured Case Study
AI-Enabled Document Review — Reading the documents nobody wants to read
Applied AIA production pipeline that reviews insurance servicing documentation — new business, cancellations, endorsements, and payments — against carrier rules, and escalates to a human wherever it isn't confident.
Underwriting risk hides in documents. Servicing paperwork arrives in no consistent schema, and every carrier applies its own rules to it, so the only way to catch a discrepancy was for a person to read every page against the right rulebook. That does not scale, and worse, it fails unevenly: attention drops on page forty in a way it doesn't on page one. The cost of a missed detail in insurance is not a bad user experience, it's a mispriced policy.
Scroll the diagram sideways to see the full pipeline →
- Confidence-gated automation — the pipeline is allowed to decline. Low-confidence results route to a person rather than being passed through with a caveat, and a randomized sample is reviewed even when confidence is high.
- Data residency by design — documents carrying PII are processed on locally hosted models. Model choice follows the data classification, not the other way round.
Evaluated against approximately 1,000 real applications before production rollout, with human reviewers comparing AI-generated findings against source documents and established review criteria. In production, evaluation continues through targeted review of low-confidence results plus randomized quality-control sampling.
Processes approximately 15,000 insurance servicing documents per month across new business, cancellations, endorsements, and payments. Approximately 85% are processed automatically, with human review concentrated on low-confidence results and randomized quality-control samples.
Typical processing time decreased from approximately 7 minutes to 2 minutes per transaction, blended across both the automated path and the work that escalates to a human.
The dedicated review team went from 10 reviewers to 5. That is review capacity redeployed, not payroll removed — the other five moved into other departments.
Claude and OpenAI APIs, AWS Bedrock and SageMaker, and locally hosted models for PII-bearing documents. AWS Textract handles structured extraction on payment transactions, where the incoming documents are standardized enough that deterministic extraction is the right tool. Python and SQL Server, on AWS and Azure.
Case Study
iBinderbook — One system of record for an insurance agency
Platform / CRMA custom agency management and CRM platform integrating policy workflows, lead tracking, and automation — designed to increase throughput and reduce manual friction.
Manual workflows and fragmented systems slowed agent velocity and introduced operational risk across policy management, follow-ups, and commission logic.
- Designed end-to-end workflow orchestration across sales → policy → servicing
- Automated repeatable tasks with clear ownership and auditability
- Integrated operational reporting for visibility into throughput and bottlenecks
650+ users on a continuously evolving operational platform spanning CRM, policies, accounting, reporting, dashboards, metrics, analytics, and integrated communications. In production since 2014 and still the system the business runs on.
Commercial agency management systems license per seat, per month. At 650 users that line runs well into six figures a year — gross, before netting what the platform costs to build and run in-house. Exact per-seat rates move over time, so the magnitude is the honest claim, not a quoted figure.
The licensing figure is the easy half of the comparison. The harder half is that owning the product means the policy, commission, and servicing workflows match how this agency actually operates, rather than how a vendor assumes it does — which is the reason the build decision held for twelve years, not the line item.
Proprietary enterprise application architecture integrating operational workflows, CRM capabilities, policy management, accounting functions, reporting and analytics, and API-driven communications. Integrated communications run on Twilio — voice, SMS, automated messaging, IVRs, and number management.
Projects
2026Active
AI-Enabled Document Review
Reviews roughly 15,000 insurance servicing documents a month against carrier rules — multimodal analysis into a structured 50+ item checklist, then deterministic business rules. About 85% clear automatically; the rest route to a human by confidence. Processing time went from ~7 minutes to ~2.
2014Active
iBinderbook
Custom agency management and CRM platform unifying policy workflows, lead tracking, commission logic, and automation into a single system of record.
2026
AI Voice Training Platform
Scenario-based employee training. Trainers define customer personas, emotional state, difficulty, behavioral rules, and success criteria; trainees hold real-time voice conversations with the simulated customer, then get AI scoring, pass/fail, strengths and weaknesses, and a persistent session history.
2023
Dealer Verification Platform
A platform I architected and own, supporting dealership employees across approximately 30 dealerships. Tracks lead status and routes insurance opportunities generated through dealership workflows.
2021
Enterprise Communications Platform
API-driven communications built on Twilio and integrated directly into the CRM platform — voice calls, SMS, automated messaging, IVRs, number management, and application-driven communication workflows.
Experience
Present
Chief Technology Officer
Insurance Nation · Miami, FL
- Set technology strategy with ownership and lead the development and IT organization
- Own build-vs-buy for core systems — chose to build the agency management platform in-house rather than adopt an off-the-shelf AMS
- Built and scaled multi-cloud infrastructure supporting every core operational platform
- Shipped AI policy-documentation review to reduce underwriting risk, with human escalation in the workflow
- Built an AI training module using insurance personas to onboard new employees
- Run cross-functional delivery across product, operations, and engineering
2019
IT Director
Insurance Nation · Miami, FL
- Launched iBinderbook in 2014 and led its expansion as the company's internally developed management platform, adding operational capabilities as business requirements evolved
- Built and expanded communications capabilities through Twilio integrations — automated SMS, voice calling, IVRs, number management, and API integration with iBinderbook
- Held technical leadership of the company's core operational systems and their integration into day-to-day business processes
2013
Systems Administrator & Business Development
Insurance Nation · Miami, FL
- Conceived the original architecture for iBinderbook and defined its requirements, laying the groundwork for the build that began in 2014
- Led early technology initiatives including the company's communications platform, establishing the foundation for broader technology leadership
Leadership & Ownership
Leading a team of 5 across development and IT, and owning the build-vs-buy decisions for every core system the agency runs on. Three that mattered:
Build the agency management system
Chose to build iBinderbook in-house instead of buying an off-the-shelf AMS. We own the product outright, and the policy, commission, and servicing workflows match how this agency actually operates rather than how a vendor assumes it does.
Automate policy documentation review
Built AI review of servicing documentation to catch what manual review misses, directly reducing underwriting risk. Roughly 85% of about 15,000 monthly documents now clear automatically; the dedicated review team halved, and those employees moved into other departments rather than out of the company.
Train people with AI, not slides
Built a voice training platform where trainers define a customer persona, emotional state, and success criteria, and trainees hold a real conversation with the simulated customer — then get scored on it. New employees practice the hard calls before taking a real one.
Education
2025
M.S. Computer Science
Georgia Institute of Technology
2008
B.E. Telecommunications Engineering
Universidad Tecnológica de La Habana, CUJAE
Certifications
Skills & Stack
Cloud
AWS and Azure — infrastructure, identity, and networking across both, with workloads placed by data-residency constraints rather than vendor preference.
Applied AI
Claude and OpenAI APIs, AWS Bedrock and SageMaker, and locally hosted models for data that can't leave. Document extraction with native multimodal models, and AWS Textract where documents are standardized enough for deterministic extraction. Confidence thresholds, human escalation, data-residency controls.
Engineering
Python, SQL Server, Twilio, REST APIs, Docker, Linux, PowerShell, Git.
Leadership
IT strategy, modernization programs, build-vs-buy ownership, governance, and stakeholder alignment.
Contact
LinkedIn is the fastest way to reach me. If you're working on document-heavy operations in a regulated industry — insurance, lending, healthcare — that's the conversation I'll always make time for.