Projects using OpenAI in Arlington HeightsProjects using OpenAI in Arlington HeightsI have built tons of Ai Agents using Openai, Claude, Gemini, and a handful of other ones. The Ai agents were voice receptionist, order takers/placers, customer support etc.
Now as I am building my 2nd app, i've come to realize that a lot more goes into training and distilling models. It can get pretty expensive if you have no idea what you are doing, as i'm doing more and learning more, i've come to the conclusion, those who integrate well with 3rd parties, I prefer to use.
Example, Supabase, Vercel AI SDK5, OpenAI, Assistant UI for UI and tooling, life can't get much easier. I drove data-backed decisions across the supply chain, product development, and executive reporting functions. Early on, I conducted comprehensive market and supplier research—building automated pricing models to benchmark vendors and calculate cost variances. When tasked with stepping up to independently manage inventory and vendors, I engineered a suite of automated forecasting tools and dashboards using Excel, Google Sheets, VLOOKUP, and BigCommerce data. These tools visualized real-time production status (W3s), predicted imminent stock-outs using color-coded heatmaps, and tracked high-volume sales. I synthesized this complex operational data into three streamlined, daily dashboards presented directly to the CEO, while also modeling emergency sourcing scenarios to ensure we maintained 100% order fulfillment during production delays. Flagspill is an anonymous social platform where users share short takes on relationship and dating situations, then vote them red flag, green flag, or contested "yellow flag." It's Reddit meets a polling app, focused on the moments people overthink and want a second opinion on.
Features:
Anonymous post and comment system with optimistic vote updates
Multi-feed browsing (Hot, New, Top Today, Top Week, All Time)
Yellow flag detection for contested posts where the community is split
Responsive masonry grid layout that adapts cleanly across devices
Real-time engagement metrics on every post
Where it gets interesting is the moderation. Flagspill runs a multi-stage moderation pipeline using regex hard rules, heuristic spam and gibberish detection, and GPT-4o-based content review with rate limiting. Every submission goes through this pipeline before hitting the database, keeping the platform safe and useful at scale.
Tech stack: React, TypeScript, Tailwind, Supabase for database and auth, OpenAI GPT-4o for moderation, deployed on Vercel.
Built and shipped solo. Live at flagspill.com (http://flagspill.com). MixerAI is a Chrome extension that consolidates ChatGPT, Claude, Gemini, and Perplexity into a single dashboard. Most heavy AI users are paying for 2-4 subscriptions and constantly switching tabs to use the right model for the right task. MixerAI fixes that.
Features:
Unified connection status for all major AI services in one panel
Smart subscription tracking so you can see which plans you're on across providers
Plan recommendation engine that suggests Free vs Pro vs Max tiers based on your usage patterns
Single task input that routes your prompt to the best-suited AI for the job
Built with TypeScript, React, Chrome Extensions API, Tailwind, and integrations with OpenAI, Anthropic, Google, and Perplexity APIs.
This project came from frustration with my own AI tool sprawl: four tabs open, two paid subscriptions I wasn't fully using, and no system for deciding which model to ask. MixerAI is the tool I built to solve that for myself, and now I'm shipping it for others. Happy Lo-Fi Holidays is a holiday-themed React web experience inspired by classic “lo-fi beats” videos and reimagined as an interactive, customizable scene.
Choose from cozy interior spaces, different exteriors, layer dynamic weather effects, and optionally mirror real-world local weather. Each scene can be paired with original Suno-generated lo-fi holiday tracks, creating a calm, immersive escape from hectic shopping and end-of-year chaos.
All music was generated with Suno using prompts focused on warm, nostalgic, winter lo-fi moods. Visual assets were created using an AI-assisted pipeline: DALL·E, Google Flow, and Adobe After Effects, enabling animated custom art.
The built-in share feature lets users send a favorite scene and song combination as a digital holiday greeting card.
Get your holiday vibe on: https://lofi.xozy.app Presenting iris- a mobile/tablet-first AI point-of-sale system built with Expo, React Native, Supabase, Roboflow, and OpenAI. It uses the device camera for real-time object detection and LLM-powered naming, letting stores build tickets or receive shipments automatically. No barcode scanner required.
The architecture is ready for cloud scaling, with clear paths to use AWS Rekognition, Lambda, DynamoDB, and S3 for enterprise-grade deployments.
**Because iris depends on mobile hardware, the web simulator can’t show the full flow, Expo Go is the preferred experience.**
Public App:
https://2b852a9b-3b94-451c-9256-457bf78e2e9c.canvases.tempo.build/
Canvas Editor:
https://app.tempo.build/canvases/2b852a9b-3b94-451c-9256-457bf78e2e9c/editor
