Instituto PROA, a Brazilian nonprofit, has revolutionized its job preparation services for young candidates by integrating Llama and Oracle Cloud Infrastructure (OCI). This AI-powered assistant automates the research process for job openings and employers, leading to a remarkable 60x growth in program enrollment—now reaching 35,000 students annually. The assistant communicates with students in Portuguese, scouring the web for valuable job insights and delivering detailed reports to support interview readiness. The seamless integration between Llama and OCI ensures smooth deployment and compatibility with PROA’s technical infrastructure.
Before adopting AI, PROA staff spent an average of 30 minutes manually compiling each report by searching multiple websites. With the new system, this time has been slashed to just five minutes per report, dramatically increasing efficiency. The automation not only saves time but also ensures consistent, high-quality information, freeing up the team to focus on personalized student support. Alini Dal Magro, Executive Director at Instituto PROA, emphasizes, “Llama is a core component of our AI-powered solution to enhance the efficiency of our job preparation process for young candidates. We chose Llama 3.1 for its seamless integration with our existing Oracle Cloud Infrastructure, enabling us to leverage OCI’s scalability and performance for AI workloads.”
The system operates through a Retrieval-Augmented Generation (RAG) architecture: when a user inputs a company name, the query is sent to a search engine API to fetch relevant web results. These results are stored in PROA’s knowledge base, where Llama 3.1 processes the data, combining retrieval with generative capabilities to produce structured, comprehensive reports. The output is automatically formatted into a PDF and shared with candidates for interview preparation. Beyond this, PROA is exploring enhancements like personalized insights in PDFs and tool-calling features to further automate candidate support.
Looking ahead, the release of Llama 3.2 opens up opportunities to leverage multimodal capabilities and lightweight models. The 11B and 90B vision models, which process both text and images, could enrich candidate dossiers by analyzing company infographics or job-related visuals, providing richer insights. Dal Magro notes, “The primary goal remains bringing more efficiency and impact to our platform, while helping low-income young people succeed in job searches with high-quality, accessible resources.”
