Solutions Architect

Juan Fazzini

I design and build production systems. I have been doing that for over 25 years, most of it in healthcare, law enforcement, legal, and fintech. I was one of the five people who started the company Making Sense and spent twenty two years there, most of it as its lead architect, before moving to fintech.

Dallas-Fort Worth · English and Spanish · CEH · CJIS Level 4

See the Work LinkedIn ↗ hello@juanfazzini.dev
01 — Work

Selected work

08 of 16 projects · Problem → Approach → Outcome
FinTech · under NDA2025 — now

Tokenized lending platform for private real estate loans

ProblemPrivate real estate lending runs on spreadsheets, wire transfers, and trust.
ApproachInvestors fund loans through tokenized vaults after identity checks. The lending team originates, acquires, and services loans through the full life cycle. A public explorer mirrors on-chain activity in real time. The servicing ledger, the smart contracts, and the investor's screen have to agree at every step.
OutcomeOne platform for investors, lenders, and the public, with an AI underwriting assistant alongside it. Infrastructure as code, encrypted key management, and audit-safe records of every privileged request.
TypeScriptNodeFastifyNext.jsPostgreSQLNATS JetStreamAWSTerraform
FinTech · under NDA2025

Multi-chain transaction platform that heals itself

ProblemA transaction has to be signed with keys nobody can steal, submitted at the right price, and confirmed even when the network is congested or the first attempt fails.
ApproachOne service orders, prices, signs with hardware-backed keys, submits, and tracks every transaction across Ethereum-compatible chains and Solana. It watches every active wallet, classifies each problem as it appears, and applies the fix on its own: resubmit, re-price, fill gaps, reorder.
OutcomeA wallet that never needs someone paged at night, and a full audit trail when compliance asks what happened. Shipped with a published SDK, operator tool, admin console, and a simulation environment.
TypeScriptgRPCPostgreSQLRedisviemSolana KitCubeSignerPrometheus
Law enforcement · COPsync2016 — 2018

FIPS 140-2 and CJIS-compliant identity server

ProblemThousands of officers across many agencies, each with its own directory and policies, needing criminal justice data in seconds from devices that could be stolen, under rules written by the FBI.
ApproachAgencies federate existing directories through Active Directory and SAML. Authentication adapts to the situation: SMS, phone, time-based codes, fingerprint, or magnetic card. One security token service issues JWT and OAuth 2 tokens; an attribute-based access control engine decides who sees what; all cryptographic material lives in hardware security modules.
OutcomeOne login for the whole product line, an audit trail regulators accepted, and a platform the company could add products to without rebuilding security each time.
.NETIdentityServerAzureKey VaultHSMSAMLOAuth 2
Healthcare · CaptureRx2017

Cumulus 4: 340B pharmacy claims as a stream

ProblemEvery 340B prescription filled at a contract pharmacy has to be matched against patient encounters, checked for eligibility, priced, and tracked through inventory and payment, with an audit trail that survives a federal review.
ApproachClaims treated as a stream rather than a nightly batch. Every claim, encounter, and inventory event lands on Kafka the moment it arrives. Spark jobs join claims to encounters, apply the 340B rules, and emit results back onto the stream. Elasticsearch indexes the history; Redis holds the hot eligibility lookups.
OutcomeInventory, financial position, and compliance status current within minutes instead of the next morning. Replaying the stream rebuilds any view, which is what makes the audit trail trustworthy.
JavaSpringKafkaSparkElasticsearchRedisPostgreSQLAWS
Legal · Esquire2023 — 2024

AI legal notice processing inside the client's cloud

ProblemEvery deposition starts with a notice in whatever format the law firm chose to send it. Schedulers were reading them and keying jobs into NetSuite by hand, at high volume.
ApproachA notice lands in secure storage and triggers a pipeline inside a private network. Text is extracted, a model on AWS Bedrock pulls out the case, parties, date, location, and requested services, and the result is written into NetSuite as a draft job. Models run behind AWS's boundary, so client documents never leave Esquire's cloud.
OutcomeSchedulers confirm drafts instead of transcribing notices. The team can switch between Anthropic, Meta, and Amazon models without changing the pipeline.
TypeScriptAWS LambdaS3BedrockTerraformNetSuite SuiteCloud
Logistics · SECO2024

Retiring an IBM AS/400 without stopping shipments

ProblemFor decades the freight consolidation operation ran on an AS/400. Every change went through a shrinking pool of people who could program it, and the business could not see its own operation in real time.
ApproachMicroservices organized around the way freight moves: receiving, transloading, consolidation, pool distribution, carrier and retailer integration. Each service owns its data and publishes events. Cutover starts with a migration phase: data cloned to both systems, new services validated against the AS/400, legacy feeds removed one at a time.
OutcomeThe business never stops shipping during the transition. Delivered the domain model, service boundaries, integration contracts, migration plan, and cloud architecture.
TypeScriptNodeAWS EKSEvent-drivenDDD
Public safety · COPsync2013 — 2018

COPsync911: real-time threat alerts for schools

ProblemWhen something goes wrong in a school, the call goes to 911, a dispatcher takes the details, then radios officers. Every handoff costs seconds.
ApproachA teacher triggers an alert from a desktop or phone; the five officers closest to the building see it on in-car and mobile screens at once, with location, alert type, and a live two-way channel. Alerts travel over multiple paths with acknowledgment, positions come from the in-car network, and clients survive network drops without losing state.
OutcomeAdopted by school districts across Texas and became the reference architecture for real-time communication across COPsync's product line.
.NETWPFSignalRWCFSQL ServerAzureMSMQ
Product · Making Sense2004 — 2016

Doppler: email marketing platform, first line of code to regional leader

ProblemSending millions of emails means managing delivery reputation with every inbox provider, throttling per destination, handling bounces automatically, and scheduling so a big customer's campaign never delays a small one.
ApproachA load-balanced, highly available web tier, a distributed sending and tracking pipeline, and a data model that scaled from the first customer to the whole region. Re-architected in 2011 for the next decade of growth.
OutcomeOne of the leading email marketing platforms in Spanish-speaking markets, still in business at fromdoppler.com. Its framework became the foundation of Making Sense's other products.
.NETASP.NETSQL ServerHigh-volume SMTP
02 — Experience

Experience

The pattern across all of them is the same. I take a business problem with real compliance constraints, turn it into an architecture the team can actually ship, and stay hands-on through delivery.

  1. 2025 — now
    Chief Solutions Architect
    A fintech company (under NDA) · Dallas-Fort Worth

    Chief architect for a fintech company bringing private real estate lending on-chain. I joined to build its blockchain infrastructure and now lead the architecture of the lending platform itself, from the first design to production.

  2. 2018 — now
    Advisory Software Architect
    DigitalBlue Software · part-time

    Technical direction and the shared foundation for public safety products: authentication server, cloud services hub, container infrastructure. EnforceNet, a FirstNet-listed safety app, and Ledexable ship on it.

  3. 2022 — 2025
    Chief Solutions Architect
    Making Sense · Dallas

    Head of architecture for a software consultancy of hundreds of engineers. Technical direction across client engagements, architecture reviews, and hands-on work on the hardest builds. Most of it AI and modernization.

  4. 2006 — 2022
    Solutions Architect
    Making Sense · San Antonio, then Dallas

    Lead architect for Making Sense's own products and its US clients, including six years embedded at COPsync and engagements at CaptureRx, Academic HealthPlans, and FCA.

  5. 2002 — 2006
    Founding team, Senior Developer and Solutions Architect
    Making Sense · Mar del Plata, Argentina

    One of the five people who started Making Sense. Wrote much of the first code, led the first client projects, and started designing Doppler in 2004.

  6. 1999 — 2003
    IT Maintenance, Junior Developer
    GIE Group · Universidad CAECE

    Software, hardware, and network maintenance; basic internal application code maintenance.

03 — About

What I bring

Over the last few years I have added AI to that toolkit: LLM retrieval over company data, agentic workflows on Bedrock and OpenAI, and AI-assisted development that produces code a human can still read and maintain.

CertifiedCertified Ethical Hacker (CEH), EC-Council
FBI CJIS Level 4
EducationComputer Software Engineering, Universidad CAECE, 2000 to 2003
LanguagesEnglish, Spanish
  • Regulated systemsGround-up architecture for regulated systems: HIPAA, CJIS, FIPS 140-2, and financial compliance
  • CybersecurityCybersecurity across the stack: identity and access management, threat modeling, secure architecture, backed by a Certified Ethical Hacker certification
  • AIAI and LLM systems that hold up in production: retrieval over company data, document pipelines on Bedrock and OpenAI, agentic workflows with a person in the loop
  • Distributed systemsDistributed systems, event-driven design, and blockchain (ledgers, wallets, smart contracts)
  • Hands-on deliveryHands-on delivery in .NET, Java, TypeScript, and Python on AWS, Azure, and GCP
  • Technical leadershipTechnical leadership: growing developers into architects and keeping teams unblocked
04 — Contact

Get in touch

Based in Dallas-Fort Worth. Bilingual English and Spanish.

hello@juanfazzini.dev
linkedin.com/in/juanfazzini