Dira App: AI-Driven Decision Support for Career Path and Employment Opportunity Recommendations: A Case Study of Tanzania
Year: 2026
Type: article
Abstract: Career and employment decisions among Tanzanian secondary school students and graduates are still largely shaped by informal advice from parents, relatives, and peers rather than by objective, data-driven guidance. Graduate unemployment in Tanzania remains high, and structured support for self-employment remains limited. This paper presents the Dira App, Dira being the Swahili word for "compass", reflecting the system's role in helping users navigate career and employment decisions. It is a web and mobile Decision Support System (DSS) that combines a machine learning Career Predictor, an AI assistant (Dira AI) that personalises guidance using each user's academic history and predictions, and a booking and video counselling subsystem for continuous, personalised support. The system was developed using Agile methodology on a Vue.js, Flask, and PostgreSQL stack. A questionnaire-based needs assessment of 114 respondents (20 secondary students and 94 graduates) confirmed the problem: 60% of graduate respondents were unemployed, only 7% were formally employed, and a combined 86% expressed willingness to pursue self-employment, while career decisions among student respondents were shown to be predominantly influenced by parents and peers. These findings directly informed the system's functional requirements, architecture, and machine learning feature set. The resulting system integrates five core subsystems: role-based User Management, ML Career Prediction, Dira AI, Booking & Video Counselling, and Reporting & Analytics behind a secure, role-based interface. The developed system successfully centralised user profiling and career/employment data, automated personalised recommendation generation through a Career Predictor trained with a Random Forest Classifier (test accuracy 94.76%, train accuracy 95.66%), and supported continuous AI-driven and human counsellor guidance for informed career decision-making. The evaluation reported here is limited to a single-institution needs assessment and functional/technical testing rather than field deployment with end users, and the predictive model was trained on data reflecting the Tanzanian context, which may limit generalisability elsewhere. By combining locally grounded predictive modelling with continuous AI-driven and human-counsellor support, the Dira App directly addresses both the local data gap and the continuity gap identified in existing career guidance literature, offering a validated, replicable decision support model for other resource-constrained East African institutions.
Authors: Ester Philipo Lulale, Alfred Kajirunga
Institution: Dar es Salaam Institute of Technology
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W7214211658