Secure login interface for accessing the MarketPulse AI competitor intelligence platform.
User registration system with MongoDB-backed authentication and persistent user data storage.
Central dashboard displaying tracked competitors, analysis status, and real-time market intelligence insights.
Live activity feed showing AI analysis events, sentiment results, and system updates.
AI-powered competitor discovery tool that finds relevant companies based on a business description and region.
Automated competitor intelligence reports summarizing sentiment, insights, and strategic recommendations.
AI strategy evaluation tool that analyzes business strategies against competitor intelligence data.

Inspiration

Have you ever tried to track what your competitors are doing? It usually involves having 50 tabs open, refreshing Google Shopping, reading news articles, and manually pasting prices into a giant, messy spreadsheet. I read a recent industry report that said 61% of companies still do this manually. By the time you finish the spreadsheet, the data is already old. We wanted to build a tool that completely automates this grunt work so teams can focus on actual strategy instead of mindless data entry.

What it does

MarketPulse AI is basically an automated market analyst that never sleeps. You type in a competitor's name, and the app goes out to the web, scrapes their live pricing and recent news, and feeds all that messy data into an LLM. Within seconds, it spits out a clean, structured dashboard showing you their pricing strategy, market sentiment, and key insights. It turns hours of Googling into a single button click.

How we built it

We stitched together three main pieces to make this pipeline work autonomously:

The Scraper: We used Apify's custom actors to go out and scrape live, real-world data from Google Search and Shopping.

The Brains: We piped that raw, unstructured data into OpenRouter. We prompted the LLM to act as a financial analyst, pulling out price anomalies and sentiment.

The Frontend: We built the user interface using Next.js, React, and Tailwind CSS so the output actually looks like a professional SaaS dashboard, and deployed it live.

Challenges we ran into

Live web data is incredibly messy. Getting the Apify scraper to return exactly what we needed was tricky, but the absolute hardest part was taming the LLM. The AI kept wanting to be "chatty" and would return conversational text instead of the strict JSON format our Next.js frontend needed. We spent hours tweaking the prompt engineering and adding validation logic just to stop the app from crashing when the LLM decided to randomly add "Here is your report!" at the beginning of the data payload.

Accomplishments that we're proud of

Honestly, we are incredibly proud of getting the whole end-to-end pipeline working autonomously in a single weekend. It’s one thing to run a script locally, but successfully getting a deployed Next.js app to trigger a cloud scraper, poll for the results, pass them to an LLM, and render the output dynamically on a live dashboard is a massive technical win for us. We built a real, functioning product.

What we learned

We learned a lot about strict prompt engineering specifically how to force an LLM to output programmatic data structures instead of prose. We also got a crash course in handling asynchronous polling (waiting for Apify to finish its run before asking OpenRouter to step in) without timing out our server.

What's next for MARKETPULSE AI

We want to take this from a dashboard to a proactive alert system. The immediate next step is hooking this up to webhooks (like Slack or Discord) so that the moment a competitor drops their price, your team gets an instant notification. We'd also love to build out direct integrations into CRMs like Salesforce or HubSpot.