Eight months at BabyLoveGrowth.ai

Summary

BabyLoveGrowth.ai is an AI SEO platform: it writes articles for a site, runs a backlink exchange and tracks how AI assistants mention a brand. Its homepage says more than 5,000 businesses use it³. I joined in February 2026 as a software engineer and work across the Next.js dashboard, the Express and Postgres backend, and new products from their first commit.

In 231 days I merged 499 pull requests across six repositories¹. Two of the larger pieces have their own write-ups: the MCP server that connects Claude to Search Console and the weekly report that failed silently twice.

Pace

What I shipped

A Quora visibility engine, then a cheaper one

Brands wanted to show up in the Quora threads that rank on Google and feed AI answers. I built the engine that finds questions for a site's keywords every three hours, scores their relevance with an LLM and drafts answers people can edit and post. Four days after launch I cut its scraping to one search call per keyword: the search results already carry the question and its answer count, so it no longer fetches a single page. The trade-off was losing each question's follower and view counts.

Official Google Ads keyword data

Keyword volumes came from a third-party proxy. I replaced it with Google Ads' own historical metrics: search volume, top-of-page bids averaged into a CPC, and a 12-month trend per keyword, behind an adapter so no caller had to change. When Google blocked the API version we used on 5 August, I checked the old and new versions against each other live and moved us to the new one.

Self-serve white-label onboarding

Agencies can now put the product on their own app domain and send email from their own domain without anyone on our side touching DNS. Domains are verified through the hosting provider's and the email provider's APIs, brand colours and the logo are read from the agency's site, and a claimed domain that is never finished is released after three days.

A Next.js blog package on npm

I wrote every commit of babylovegrowth-next-js-blog: a dependency-free client and an init command that scaffolds a complete blog into a customer's Next.js App Router site, with list, article and tag pages, a sitemap, JSON-LD and daily revalidation².

Business profiles that stop off-target articles

Articles were inferred from a site's niche alone, and customers kept reporting that they described services the business does not offer. I added extraction of services, price ranges and positioning to the crawl the platform already runs, at no extra cost, plus a settings page to correct it and a backfill job. Article generation and the claim checker both read it now.

Smaller fixes worth naming

• Facebook and Instagram visits were counted as AI search traffic, because the source “meta” fell through to the Meta AI rule, and substring matching turned unrelated hosts into You.com. Matching on the hostname fixed both.
• Parallel onboarding jobs hit a search API's rate limit, and it answered with 200 and empty results instead of an error. Requests now self-throttle, and an empty result is treated as a retryable failure, not as an answer.
• Four free SEO tools on the public site: an H1 checker, a SERP snippet preview, a Google index checker and a robots.txt generator.
• Italian as the seventh dashboard language, generated from a single locale list.