Building intelligent automation solutions that streamline workflows and eliminate inefficiencies
📍 Greater NY/NJ metropolitan area
Nine years in operations taught me one thing: most business problems are process problems. I've managed $20M+ in inventory, run crews across four states, and cut delivery failures by 75% — all by finding where the system breaks and fixing it.
Now I do the same thing with AI. After earning my AI Automation certification, I build automated workflows that eliminate the manual bottlenecks I used to fight every day. Make.com, UiPath, API integrations — these are just better tools for the same job I've always done: making operations actually work.
Most people automating business processes have never been inside a broken one. I've spent nine years there. That's the difference.
First-responder supply chain at Air Brake & Equipment
Client acquisition and operational scaling at Maverick
Multi-state installation teams across the tri-state area
End-to-end workflows built with Make, UiPath, and AI
Explore a collection of innovative AI automation projects that have transformed business operations, improved efficiency, and delivered measurable results across various industries.
Problem: Inventory discrepancies were resolved by hand — a process that took 2-3 months and an hour of search time per missing item.
Solution: Built an AI-powered OCR pipeline that processes pick tickets and packing slips in real time.
Eliminated 2-3 month processing delays, reduced warehouse search time by 1 hour per search, improved customer satisfaction
AI-powered OCR processes pick tickets & packing slips
Confidence scoring routes high/low accuracy items
Instant inventory adjustments via API integration
Automated notifications for low stock levels
Transformed inventory management from a manual, time-intensive process into a fully automated, real-time system that eliminated months of processing delays, dramatically reduced warehouse search time, and significantly improved customer satisfaction through accurate, instant inventory tracking.
Problem: Invoice processing was manual, slow, and prone to data entry errors.
Solution: Built a UiPath RPA solution that handles the entire workflow — document processing and data entry — automatically.
90% reduction in manual processing time, 99.9% accuracy
Problem: An ice cream and coffee business was making staffing and inventory calls based on season, not actual conditions.
Solution: Built an integration that feeds live weather data into daily operational decisions.
80% improved decision-making through dynamic weather-based operations
Problem: Clients needed professional landing pages fast — traditional development was too slow.
Solution: Built a rapid deployment system on Lovable that cuts launch time by 70%.
Problem: Customer feedback was arriving but nothing was routing it, tracking it, or responding to it consistently.
Solution: Built an automated pipeline that handles collection, processing, and follow-up without manual intervention.
Streamlined feedback collection and response workflows
Problem: Hotel staff were handling repetitive guest questions manually, pulling time away from higher-value work.
Solution: Built an AI chatbot that handles common inquiries automatically with a 95% success rate.
95% success rate, reduced staff workload
Problem: Process mapping was a slow, manual exercise — teams spent hours diagramming workflows before any analysis could begin.
Solution: Integrated AI into Miro to accelerate design and surface insights faster.
Streamlined workflow design and analysis processes
Ready to transform your business operations with AI automation?
Let's Discuss Your ProjectWe ran 4 AI models on the same real ops data — summarization, priority extraction, JSON output, and message drafting. Same prompt. Every model. Here's what came back.
Write a 3-sentence summary of what needs attention this week.
Open items for the week: - Follow up with client on proposal sent Tuesday - Review Q1 budget draft before Thursday meeting - Respond to 3 support tickets marked urgent - Update project timeline — milestone 2 slipped by 4 days - Schedule onboarding call for new team member starting Monday
Run it yourself:
python3 llm-benchmark.py --tasks summarizeAccurate and well-structured. For sensitive operational data, it's a non-starter — and even when the data probably isn't sensitive, we'd rather not find out the hard way.
Run it yourself:
python3 llm-benchmark.py --tasks summarizeFast and cheap. Output quality held up on simple tasks but drifted on the more complex ones.
Run it yourself:
python3 llm-benchmark.py --models gemma --tasks summarize --skip-claudeMost consistent across all four task types. Ran on a $1,100 Mac Mini — no cloud, no subscription, no data leaving the building. This is the one we deployed.
Run it yourself:
python3 llm-benchmark.py --models mistral --tasks summarize --skip-claudeCapable model — output quality is version and size dependent. The 7B we ran was slower and less consistent than Gemma on structured tasks.
Full results across all 4 tasks — summarize, priorities, classify, draft — in the sample report and article series below.
Interactive experiences built with modern web technologies. Combining creativity with technical skills to create engaging digital games.

Created for my nephew Roddy, this digital adaptation brings a beloved physical board game to life. Built using an AI-first development approach: parallel PRDs generated by Claude and ChatGPT, both prototyped in Lovable, with the winning version refined in VS Code and deployed via Harness CI/CD pipeline.
Validated AI PRD methodology comparing Claude vs ChatGPT outputs. Established rapid prototyping pipeline using Lovable for quick iteration. Built production-ready game with comprehensive test suite via Harness.
Explore my latest writings on AI automation, operations management, and technology innovation on Medium.
The right tool for the right job.We tested 4 AI models on real business data. Here’s why we chose the local one. A lot of AI comparison articles you’...
The whole thing fits in a file you can read in two minutes.Simple automation doesn’t need to be complicated. When people hear “AI-powered standup bot...
The summary was already there when the team arrived.No cloud. No subscription. No data leaving the building. The question wasn’t whether AI could run...
Your data stays on the desk, not in the cloud.When your automation depends on someone else’s server, you’ve already lost control. Every time you ente...
🏆AI Comparison — ChatGPT, DALL-E, and Nano Banana: Round 2 I returned to AI image generating platforms for a new project — and this time ran the sam...
Most AI automation projects fail. Not because the technology is bad — but because the people building them have never done the job they’re automating...