On April 4–7, EffectiveSoft hosted its second hackathon centered around Generative AI. This time again, the event brought together professionals from across the company—developers, QA engineers, designers, analysts, and others. For several days, they worked side by side, brainstorming, eating pizza, learning from each other, and building working prototypes.
Projects and results
Each team took a different approach to applying Generative AI to real-world challenges. Here’s what they created.
Lead scoring system
The team that eventually won the hackathon built a multi-agent pipeline based on LangGraph and GPT-4o that processes and validates visitor activity and interprets behavioral buying signals. The system then sorts analyzed leads based on their overall scores and creates a structured output—ranked lists, a summary report, and personalized email drafts for top leads. All data is handled securely, with the utmost respect for privacy, and in full compliance with regulations.
Talent search assistant
The AI assistant helps automate candidate screening. It extracts key information like skills, roles, and experience from CVs and stores it in a structured format. This solution allows recruiters to quickly match candidates to job descriptions and get a detailed analysis of selected profiles, which significantly speeds up the hiring process.
Candidate evaluation system
While similar to the previous project, this assistant focuses more on evaluation of the candidates. It compares profiles semantically, highlights key gaps or strengths, and ranks candidates based on skills, experience, and fit, also providing reporting to track hiring trends and vacancy performance over time.
Multi-agent system for fleet analysis platform
The team created a prototype of an AI solution within an existing client project for a fleet analysis system. Behind the scenes are multiple agents that extract data from unstructured and semi-structured documents, assist with platform navigation, and build module-specific charts and tables. A workspace builder agent orchestrates them all, allowing users to generate custom workspaces. With both text and voice input, interacting with the platform becomes more intuitive.
Lead research tool
The tool helps sales teams get relevant information about potential clients—company details, location, and contacts—from trusted sources like Crunchbase, Forbes, and TechCrunch. To make answers even more accurate, the assistant can rewrite and improve the original query. It also uses a retrieval-augmented generation approach to minimize external requests and ensure results are based on the most recent data.
Looking ahead
The ideas born over these few days are already sparking new conversations, experiments, and plans for what’s next. We’re excited to see where these projects lead, and we are already looking forward to the next hackathon.
