News & Tender Intelligence Portal

Business Problem

R&D and Marketing teams previously had to manually search multiple sources for relevant agriculture and industry news, research developments and government tenders. That information was fragmented across different sources, which made monitoring time-consuming and difficult to consolidate into anything usable day to day.

Automation Opportunity

I identified the opportunity to replace that manual, multi-source searching with a centralized intelligence portal that automatically collects relevant information and presents it in one place — built around configurable keywords rather than a fixed, one-size-fits-all feed. Current keyword examples include seaweed, sagarika, biostimulant, seaweed powder and carrageenan; users can add or remove keywords as their information requirements change.

Solution Approach

I developed the concept and solution using AI-assisted development with Claude. The portal consolidates relevant news and tender information from multiple sources: Python handles the scraping and collection workflow, using BeautifulSoup for web scraping and RSS feeds for news collection. Because the user base is small, the portal currently runs locally — R&D and Marketing users run the application themselves and work with the consolidated results directly, rather than this being a centrally hosted, always-on service.

The view below is a recreation of the portal with synthetic data — branding anonymized and article content replaced with representative examples.

Key Responsibilities

• Identifying the business problem and proposing the centralized intelligence solution
• Designing the concept and desired workflow
• Using Claude as an AI-assisted development tool to build the application
• Working on and implementing the web-scraping workflow
• Using Python and BeautifulSoup for scraping
• Working with RSS feeds and selected news/tender sources
• Configuring the keyword-based monitoring approach
• Sharing the portal with users and explaining how to use it
• Incorporating practical requirements around keywords and information sources
• Solving the underlying business problem of fragmented manual monitoring

Technologies I Personally Used

AI Workflow

The portal is organized around three views — News, Tenders and Trends — so users land on a structure that matches how they already think about the information, rather than one undifferentiated feed. From there, users can search results, filter by keyword or time period, sort what they're looking at, and refresh the underlying data on demand. Matching keywords are highlighted directly within results, and summary cards surface counts — number of news articles, tenders, recent items and keyword hits — so a user can tell at a glance whether there's anything new worth reading. Relevant tenders link directly to their source PDFs, and users can share the consolidated intelligence onward by email.

Claude was used heavily as an AI-assisted development tool to build this workflow. My contribution was the business idea, the requirements, the keyword-driven workflow design, and working through the generated implementation — including the scraping logic — rather than writing a production backend from scratch.

Business Value

The portal consolidates previously fragmented monitoring into one place, reducing the need for users to manually search multiple sources every time. It gives R&D and Marketing a more structured daily monitoring workflow, makes relevant information easier to search and review, and lets users share intelligence with managers or leadership directly by email instead of compiling it manually first.

Lessons Learned

• The most important lesson was that the value came from solving an information-fragmentation problem, not from the scraping technology itself
• Defining the right business keywords and information requirements is critical for an intelligence-monitoring solution — the system is only useful when the collected information is actually relevant to the users
• Designing around the user's actual workflow matters: users need to quickly search, filter, review and share relevant information rather than simply receive a large volume of scraped data
• AI-assisted development can significantly accelerate turning a business idea into a working internal tool, while still requiring human judgment to define requirements, validate outputs and refine the solution
• Working with multiple external sources highlighted the importance of source reliability, data quality, and maintaining the monitoring workflow as sources and requirements change

Future Roadmap

These are proposed future improvements, not current capabilities:

• Formal internal deployment if adoption expands beyond the current limited user base
• Expanding and maintaining the number of relevant news and tender sources
• More systematic source management, so new sources can be added and maintained without ad-hoc changes
• Stronger historical storage, so users can analyze trends over longer periods rather than only reviewing currently collected information
• Automated alerts or digests for high-priority keywords or relevant opportunities
• AI-assisted summarization and prioritization, so users can quickly understand why an article or tender may be relevant
• A more scalable data layer, if the number of users, sources and historical records grows substantially

Reflection

What I learned: the project reinforced that AI-assisted development can dramatically reduce the time required to turn an idea into a working product, but the quality of the result still depends on the person's ability to understand the business problem, define requirements, validate outputs and continuously improve the workflow.

What I would improve today: I would think about scalability and information quality earlier, rather than focusing primarily on getting the first working version running. I'd design the source-management and historical-data approach more systematically from the start, with an eye toward how the solution could evolve from a small internal utility into a more formal intelligence platform.

How I would approach this differently today: I would explore AI-based summarization and relevance scoring, so users don't have to manually read every collected item. The goal would be to move from simply collecting information to helping users understand which information actually matters.