Principal software engineer · Independent consultant Central Florida · Remote, US

AI agents and production systems for businesses that run on real workflows.

Twenty-five years of software engineering. I design and build agentic AI systems, LLM-powered automation, and the full-stack products and cloud infrastructure they run on, for regulated, data-heavy, and legacy-bound environments where a demo isn't enough.

Start a conversation See selected work Taking new engagements
James Russo James Russo as a boy at a home computer
James RussoWhere it started
Why me

Three things you won't get from an agency.

Zero to production, alone.

I've taken a SaaS product from a founder's idea through data model, API, web app, multi-account AWS infrastructure, and CI/CD to paying use, and kept it shipping for two years.

AI at production scale.

Principal engineer at Overjet (2019–2022), building FDA-cleared computer-vision AI used in clinical decisions. Two issued patents in autonomous device networking; co-inventor on a published application for ML-driven claims processing. Today I build agentic systems on Claude, OpenAI, and Bedrock.

An operator's perspective.

I've run the business and technology side of a two-doctor medical practice for twenty years alongside my engineering work. I build for the people who have to use the software every day.

Services

What I take on.

Architecture, hands-on build, or a fractional principal-engineer seat.

Agentic AI & LLM workflows

Agents that do real work: process documents, reconcile records, operate other software, talk to customers. MCP server design, tool-calling architectures, retrieval pipelines, LLM-as-judge evaluation, and computer-use agents that drive legacy Windows and web applications. Unattended browser automation against portals with MFA, rotating credentials, and no API. Deployable behind a BAA or inside a private VPC when compliance requires it.

ClaudeBedrockOpenAIMCPComputer usePlaywright

Full-stack product engineering

Complete web applications from schema to screen: FastAPI backends, Angular or React front ends, Auth0 and SSO, multi-tenant data models, file and image pipelines, transactional email. Built with tests, migrations, and CI/CD from day one so the product keeps evolving after launch.

Python / FastAPIRailsAngularReactPostgreSQLAuth0

Cloud infrastructure & integration

AWS organizations and landing zones in Terraform and Terragrunt; ECS/Fargate, RDS, S3, SES, EventBridge, AppStream. Kubernetes with GitOps, Proxmox and ZFS, secure networking, DR and tiered backup. Integration layers over systems never designed for it: on-prem databases, vendor apps, accounting platforms, telephony.

AWSTerraformK3s / ArgoCDProxmoxFortiGate / UniFi

Data engineering & analytics

Warehouses and dbt models built from operational databases; bitemporal and event-sourced designs for auditability; financial reconciliation and variance detection; database migrations such as MongoDB to PostgreSQL.

dbtPostgreSQLEvent sourcingReconciliation
Selected work

Engineering summaries, not marketing tiles.

Client names withheld where the work is internal.

Venue & sports marketing · SaaS · 2024–present

Planvue

Lead engineer on a multi-tenant platform where venue operators catalog every signage and branding placement on an interactive map: drawn areas, per-location print specs (dimensions, bleed, materials, designer notes), reference photos, downloadable design templates, favorites, and per-venue user management.

Angular 17LeafletAuth0FastAPISQLAlchemyPostgreSQLTerraform / TerragruntFargateGitHub ActionsSentry
Role
Architecture, backend, front end, infrastructure, CI/CD
Data
Migrated a legacy MongoDB dataset into a relational model with Alembic migrations
Cloud
Multi-account AWS organization (shared services, dev, prod) built from reusable modules
Scale
10+ venues with interactive maps, and thousands of individual brandable locations detailed and available for selection
Healthcare · Practice-management platform · 2025–present

DSO operations platform

Principal engineer (contract) on the internal platform a mid-sized dental DSO runs its offices on. The biggest piece is insurance-verification automation: eligibility requests are generated from the schedule, routed by payer to an EDI 270/271 clearinghouse or a headless-browser scraper for that payer's portal, cross-checked, classified (payer answer vs. our failure), and written back into OpenDental, with plain-language failure reasons for the front desk and a Datadog dashboard for the RCM team. The scrapers run through a centralized egress gateway I designed so each office's traffic leaves from its own clinic network, with per-location credentials resolved from 1Password, MFA/OTP handling, circuit breakers, failure capture for post-mortems, and an LLM judge that grades ambiguous outcomes.

Around that: a merge train that batches the volume of AI-accelerated pull requests into hourly runs instead of running the pipeline on every merge, saving about $10,000 a month in CI/CD costs; migration of Rails encrypted credentials to environment-sourced secrets; DB-backed feature flags with percentage rollouts; and an OAuth 2.0 + PKCE layer so Claude and other LLM clients can reach the platform's MCP tools.

RailsSidekiqEDI 270/271PlaywrightTailscale / SOCKS51PasswordDatadogTerraformECS FargateMCP / OAuth
Desktop
Electron client that bridges OpenDental, streamed from the cloud through AppStream, and the user's own computer. When OpenDental needs a file, an SQS message has the desktop app open a native file picker and send the file to S3 for OpenDental to use; downloads run the other way. That bridge is what lets OpenDental serve nearly 70 offices across nearly 10 regions. I also moved the desktop app to the company's single sign-on: each user signs in and receives temporary AWS credentials of their own, so every SQS message and S3 transfer is tied to an individual person.
OpenDental
Maintained a customized C# fork: appointment-indicator API, referral flows, off-UI-thread image loading, tag-stamped builds with SSM deploys and rollback.
Infra
Terraform for production, utility, and marketing stacks: GuardDuty, cost-anomaly alerts, 55 CloudFront sites imported, Datadog monitors as code, Fargate Spot tracking.
Scale
Nearly 30,000 insurance verifications automated each month across nearly 70 offices
Healthcare · Practice operations · 2023–present

Dental Nexus

A bespoke platform for the 20-plus-year pediatric dental practice I help manage, built and run by me as its sole engineer. Because I work inside the practice, the loop is short: a problem spotted at the front desk on Monday can have an automated fix rolled out by Tuesday morning.

It started as insurance verification: browser automation built to run unattended signs in to insurer portals, handles one-time passcodes, matches the patient, and brings eligibility, estimates, and claims back alongside the day's schedule. It grew from there: live call transcription with AI coaching and post-call summaries; an assistant that answers staff questions from appointments and records; a task board fed by email and fax; daily deposits pushed to the accounting system, with bank transactions matched to insurance payments; and replies to Google reviews drafted by AI and approved by a person.

FastAPICeleryPostgreSQLRedisAngular 19PlaywrightGoBedrockK3sProxmoxArgoCDGitHub ActionsSentry
Role
Architecture, backend, front end, infrastructure, CI/CD
AI
Models on Amazon Bedrock with prompt caching, versioned prompts managed from an admin screen, and an evaluation harness for call summaries
Integrations
Practice-management system, insurer portals, phone system, accounting, card payments, Google reviews
Infra
Private cloud inside the office: K3s on a three-node Proxmox cluster, with ArgoCD deploying every change automatically from Git. Events between services go through a transactional outbox, so none are lost.
Testing
~2,000 automated tests
Healthcare AI · Windows · 2020–2022

Practice data and x-ray sync client

Lead engineer on the Windows client Overjet installs in dental practices to feed its AI. It extracts practice-management data on a schedule or on demand, across the major practice-management systems and five database engines, and ships it encrypted to the cloud.

X-rays move in near real time: the client watches the file system for new images, waits for the imaging software to finish writing each one, looks up the patient and exam in the practice's own databases, and uploads the image linked to the right record.

C#.NETWindows servicesPythonGoogle CloudPub/SubTWAINWebSockets
Role
Lead engineer; designed and built the client and its configuration service
Fleet
Two Windows services that supervise and update each other and take commands remotely, with signed installers and release channels
Config
Central, layered settings: global defaults, then system and version, then group, then practice. One change reaches every practice that inherits it.
Capture
Prototype for taking x-rays in the browser: a native Windows service drives the sensor through TWAIN and talks to the page over WebSockets
Telephony · Speech AI

Real-time speech pipeline

Live transcription and agent coaching for dental office phone calls. Call audio streams from Asterisk/FreePBX through Deepgram, tuned with dental and insurance vocabulary, into Redis and FastAPI, so the transcript appears on the agent's screen as the conversation happens.

While the call is in progress, the running transcript goes to a model on Amazon Bedrock that coaches the agent on what is actually being said: what to ask next, what to offer, what was missed. The coaching instructions live in prompt templates cached on Bedrock, so each turn only processes the new stretch of conversation, which keeps responses fast enough to be useful mid-call and keeps cost per call low.

After hangup, the full transcript is processed again to produce a call summary and the follow-up items that came out of it, generated automatically instead of written up by the agent from memory.

AsteriskDeepgramBedrockPrompt cachingRedisFastAPI
Healthcare · iOS · Swift

Chairside photo capture for iPad and iPhone

Native SwiftUI app for clinical staff: pick a patient from the day's schedule, photograph at the chair, review the shots, and upload them to the patient's chart in the practice-management system. Images are resized on the device before upload, the session token lives in the Keychain, and a demo mode backed by fixture data lets the app be reviewed without touching real records. Builds are archived, signed, and sent to TestFlight from GitHub Actions, with crash reporting in Sentry.

SwiftSwiftUIiOS 17KeychainTestFlightGitHub ActionsSentry
Finance ops · Reconciliation

Payment reconciliation pipeline

Automated matching of card payments recorded in the practice-management system against the card processor's own transactions: first by authorization code, then near-matches on the code, then by date. Whatever is left unmatched on either side is listed for review each month instead of buried in a spreadsheet.

Pythonpandas
Analytics · Data engineering

Operational data warehouse

Bitemporal, append-only event model over twenty years of transactional history; dbt star schema for reporting; fee-variance exception reports that flag carrier and coverage misconfiguration.

dbtPostgreSQLBitemporal
Finance ops · Integration

Accounting integration

Daily deposits from the practice-management system pushed into Xero as bank transactions, with OAuth 2.0 sign-in, token refresh, and organization selection.

Xero APIOAuth 2.0
Consumer hardware · IoT · iOS

Hardware product, inception to delivery

Founded and ran Aeroluxmaps (2020–2025), designing and manufacturing custom aviation charts that show live weather: addressable LEDs behind the chart, with fiber optics carrying each light through the map face at its airport. Electronics, fabrication, and the business around them.

The software is mine too. A Raspberry Pi in each map runs a Python application that pulls aviation weather reports and colors every airport by flight conditions. Each unit joins an AWS IoT fleet with its own certificate, issued from a certificate authority I set up, and takes configuration changes and over-the-air updates through a serverless backend.

Setup is a native iOS app in Swift, using a Bluetooth Low Energy protocol I designed: the app has the map scan for Wi-Fi networks, sends the chosen network and password, and follows the map's progress until it is online.

ElectronicsFabricationRaspberry PiPythonSwiftCoreBluetoothAWS IoTServerless
Experience

Where the years went.

2007 – present

President, Halo3 Consulting, LLC

Custom software, systems integration, cloud infrastructure, and AI engineering for small and mid-sized businesses.

2025 – present

Principal Software Engineer (Contract), mid-sized dental DSO

Principal engineer on the DSO's internal practice-management platform: insurance-verification automation (EDI 270/271 plus payer-portal scrapers), the Electron desktop client, a customized OpenDental fork, an MCP/OAuth layer for LLM access, and the Terraform, CI, and deploy pipeline behind it.

2006 – present

Business manager and co-owner, pediatric dental practice

Operations, finance, IT, and analytics for a two-doctor practice in Central Florida.

2019 – 2022

Principal Software Engineer, Overjet

FDA-cleared dental AI. Led the Windows sync client that brings practice data and x-rays from dental offices into the AI pipeline, the layered configuration service that manages the fleet, and a prototype for capturing x-rays from sensors in the browser. C#, Python, Google Cloud Platform. Remote.

2020 – 2025

President, Aeroluxmaps

Founded and ran a company producing custom fiber-optic-lit aviation maps.

2013

Software Developer, Yodle

Web platform engineering.

Patents & publications

Named inventor.

US 8,631,063Issued Jan 14, 2014

Modular platform enabling heterogeneous devices, sensors and actuators to integrate automatically into heterogeneous networks

A hardware platform with drivers that turn every attached sensor or actuator into a programmable software service behind a common middleware interface.

Google Patents ↗
US 7,895,257Issued Feb 22, 2011

Modular platform enabling heterogeneous devices, sensors and actuators to integrate automatically into heterogeneous networks

A self-describing sensor network: nodes carry their own communication software, so new devices are discovered and integrated automatically.

Google Patents ↗
US 2023/0008788Published application · Overjet · co-inventor

Point of Care Claim Processing System and Method

A machine-learning system that processes clinical image data and patient records at the point of care to adjudicate insurance claims in real time.

Google Patents ↗
About

Background and how I work.

Background

I started programming at nine, in BASIC and then Pascal, and by thirteen I was running a bulletin board system, one of the dial-up online communities people used before the web, on a copy of Telegard I had customized with features of my own. During the web-hosting boom of the late 1990s and early 2000s, I wrote Apache modules while earning my degrees in computer engineering. I was the kid who lived in Radio Shack and took everything apart to see how it worked, then tried to make it better. That is still how I approach a problem: understand how it really works, then fix the part that matters.

Away from the keyboard, I'm an instrument-rated private pilot with a lifelong interest in aviation, which is what led to Aeroluxmaps.

How I work

Architecture reviews

Fixed scope. A written assessment of your system or plan, with specific recommendations you can act on with or without me.

Build engagements

Hands-on delivery of a defined piece of work: an agent, a product, an integration, an infrastructure build-out.

Fractional principal engineer

A standing seat on your team for technical direction, hard problems, and mentoring, at a fraction of a full-time hire.

Remote-first and US-based. Comfortable working directly with founders and non-technical owners as well as engineering teams.

Education

2003 – 2005M.S. Computer Engineering, University of FloridaThesis: a self-describing wireless sensor network built with Java and OSGi
1998 – 2002B.S. Computer Engineering, Florida Atlantic University

Industries

SaaS platformsSports & venue marketingHealthcare & HIPAAFinancial operationsIoT & embeddedTelephonyAviation
Contact

Start a conversation.

Tell me what you're trying to build and roughly when you need it. I reply to every serious inquiry, usually within a business day.

No newsletter, no follow-up sequence. Just a reply.