@mayank-10242Full Stack Automation Engineer, Independent
I'm Mayank, an AI automation engineer in Boston finishing my M.S. at Northeastern, with nearly 4 years of full-stack and integration Work.
🛠️Here's what I'm Building
- AI Marketing Ad Review: rule gates screen each ad at zero AI cost, Claude returns a strict-JSON approve / review / block decision, and deterministic fallbacks take over when the model's answer is unusable
- Applications Made Easy: a self-hosted n8n job aggregator over 16 sources
- Next, the Confidently Wrong Lab: measuring how often an LLM judge is confidently wrong, and which oversight actually catches it
📈 What's working
- 834 requests, 0 failures, $0.0044 per ad, 6.4 s median
- Never trusting the model alone: the cheap rules and fallbacks around Claude matter as much as the prompt
- Sharing free templates as an n8n Verified Creator
🙋 One thing I need
A full-time AI engineering or forward-deployed role starting January 2027 (Boston, NYC, or remote). If your team hires builders who ship LLM systems that hold up in production, I'd love an intro!
Introductions
Comments2
Is AI Marketing Ad Review something you plan to sell to agencies, or mainly a portfolio piece for your January 2027 job search?
Thanks, Ryan! The plan is to turn it into a product for agencies. Right now I'm making it more robust: testing it on larger, real ad sets, adding a self-serve review form so non-technical marketers can use it directly, and building code-level checks for the moments when the AI contradicts its own findings (it did, twice, in my last test run). Once it holds up at agency volume, I'll package it for them. If any agencies here review a lot of AI-generated ads, I'd love to hear what slows down your approval process!
2 comments
Is AI Marketing Ad Review something you plan to sell to agencies, or mainly a portfolio piece for your January 2027 job search?
Thanks, Ryan! The plan is to turn it into a product for agencies. Right now I'm making it more robust: testing it on larger, real ad sets, adding a self-serve review form so non-technical marketers can use it directly, and building code-level checks for the moments when the AI contradicts its own findings (it did, twice, in my last test run). Once it holds up at agency volume, I'll package it for them. If any agencies here review a lot of AI-generated ads, I'd love to hear what slows down your approval process!