Frankie Ottomanelli
Senior Full Stack Engineer. Product engineering & AI tooling.
I ship product features for health tech and build internal AI tooling that saves engineering teams hours per week. Currently at Grow Therapy, open to collaboration.
The short version.
Self-taught engineer with 10+ years shipping software across health tech, gaming, eCommerce, and fintech, from J.Crew's checkout to Grand Theft Auto's in-game UI at Rockstar. Today I'm a Senior Full Stack Developer at Grow Therapy, where I ship product features for the mental health platform and pioneered an internal AI Developer Experience initiative — building autonomous agents that investigate CI failures, review PRs with full-repo context, and take repetitive work off engineering teams. A lot of that centers on generating repo-specific context files that ground AI in how a large codebase actually works. On the side I'm composing agentic skills (MTA trip planning, batch notifications, todo and calendar) into a personal assistant.
On the full-stack side I've rebuilt invoice-validation pipelines that cut insurance-claim denials by 36%, designed real-time dashboards with anomaly alerting, and led security hardening with end-to-end encrypted form submissions. Outside work I build small things in public: Turtto (a free ETF comparison platform, ~1,000 MAU) and an automated Prosper lending strategy that's averaged 7% annualized since 2022. Off the keyboard I'm playing basketball, teaching myself piano, and raising my daughter Nova with my wife Anna, our dog, and our two cats.
Selected work.
09/2024 - Present
Grow Therapy
Senior Full Stack Developer
Designed and led development of a session time edit request workflow and admin dashboard that centralized approval data, created the audit trail required for payor compliance, and simplified the correction process for providers, reducing support turnaround from multiple days to hours. Rebuilt invoice-submission validation that cut insurance-claim denials by 36% (11% → 7%). Shipped real-time dashboards with anomaly alerting that shrank detection from hours to minutes. Built GAD-7/PHQ-9-based 'Nudge' scoring to trigger timely therapist notifications, and secured form submissions with encryption (including migration of unencrypted data) and audited decryption workflows for the data team. From Feb–Jun 2026, also pioneered Grow's AI Developer Experience initiative — partnering with engineering and product leadership on the operating model for AI in development workflows, and shipping internal tooling: CI Investigator (an autonomous agent that diagnoses CI failures and posts root-cause summaries to Slack and PRs, saving engineers ~2 hours per week), an AI PR reviewer that evaluates changes with full-repo context, and the grow-scripts convention pairing deterministic scripts with LLM orchestration.
04/2023 - 09/2024
Seeds Investor
Senior Full Stack Developer
Led a full-stack codebase audit that cut response times by up to 75% and eliminated all high-severity vulnerabilities. Designed multi-account support enabling aggregated and account-level portfolio views with weighted-value calculations across profiles, individual accounts, and total balances. Built robust schemas and REST endpoints using dependency injection for maintainability and testing.
11/2019 - 04/2023
Rockstar Games
Full Stack Developer
Shipped internal C#/TypeScript applications for major Rockstar titles (including Grand Theft Auto), enhancing player progression and economy features and improving player engagement. Redesigned the player-engagement app around real-time analytics to better time in-game marketing campaigns. Built a shared TypeScript/React component library and consolidated SQL databases to unify cross-platform analytics.
08/2015 - 11/2019
J.Crew
Jr Developer → Tech Lead
Led the modernization of browsing and checkout on jcrew.com, refactoring legacy code and introducing React, Next.js, and GraphQL. Managed a team of three to overhaul global components and built a bespoke CMS for content management. Drove performance work (caching, image optimization, lazy loading) that measurably improved page load and user satisfaction.
Side projects.
05/2026 - Present
Stathead Natural Language Query Tool
Developer
Built a natural-language tool that translates plain-English baseball questions into Stathead query URLs using a two-stage LLM pipeline, manually curated query-parameter metadata, and a hand-graded evaluation suite; validated prototype performance.
04/2026 - Present
MTA Commuter
Developer
OpenClaw skill that answers natural-language commuter questions ('next train from Penn to Huntington', 'fastest way home by 9') across LIRR, Metro-North, and NYC Subway. Combines all three systems into multi-leg trip plans with live delays, service alerts, and track-assignment watches via the MTA's public GTFS feeds (no API key required).
06/2022 - Present
Prosper Automated Investing
Developer
Built an automated investing tool in Node.js for Prosper loans that evaluates nearly 200 data points to automate loan selection and reinvestment; the private strategy built on top of it has generated a 7% annualized return since June 2022.
03/2023 - 09/2024
Turtto
Founder / Developer
Built and launched a free ETF comparison platform that enabled users to analyze fund overlap and weighted portfolio exposure across holdings, reaching approximately 1,000 monthly unique users.
What I can build with you.
I help engineering teams ship faster. Four areas in particular:
i.
Coding Agents
Autonomous agents that work in your codebase: PR review with full-repo context; CI triage that diagnoses failures, opens draft fix PRs when the cause is mechanical, and quarantines flaky tests when it isn't; and repo automation for the maintenance work nobody enjoys doing manually. The hard part is grounding them in how your codebase actually works, which is most of what I do.
ii.
Workflow Agents
Background agents that watch queues, run on schedules, and route work between tools. Example: a watcher that polls Jira for tickets with a specific label and kicks off a coding agent to draft the PR. The real design question is which steps should be deterministic workflows with the LLM as orchestrator, and which should let the agent improvise. Get that call right and the agent just works.
iii.
Operational Agents
Agents that respond to production state. Post-deploy monitors that correlate fresh errors with what just shipped, incident triage that pulls logs and proposes a fix, anomaly investigation across metrics and traces.
iv.
Full-Stack Build & Rebuild
Production-grade full-stack apps built or rebuilt with modular architecture, performance budgets that hold under load, and a lean stack that scales without becoming a maintenance burden.
Let's talk.
The best work I've done started with a short conversation. If you're hiring a senior engineer, kicking off an AI agent project, or just want to compare notes on developer experience and AI tooling, I'd like to hear from you.
Socials