In the Democratic Republic of Congo, skilled tradespeople often lack the documentation to enter the formal job market. Oscar Mulume, a developer based in Goma, identified a critical gap: electricians, mechanics, and phone repairers possess valuable skills but cannot translate their on-the-job experience into CVs written in French or English. His solution, Mosala (Lingala for 'work'), is a local-first application that converts spoken or typed narratives in African languages into professional resumes, leveraging the power of small, offline large language models.

Architecture and Local-First Philosophy

Mosala runs entirely on the user's machine, bypassing the need for expensive cloud APIs or reliable internet connections. The stack is lightweight and builder-friendly: users install Ollama, pull a small model such as qwen2.5:3b, and execute python app.py. This approach ensures privacy and accessibility in regions where data costs are high and connectivity is intermittent. The tool specifically targets the 'slow, boring part' of job applications, automating the translation of trade-specific slang into standardized job titlesβ€”turning 'fundi wa simu' into 'Phone Repair Technician' for recruiter search algorithms.

Features Beyond Simple Translation

The application offers more than just CV creation. It includes a 'Match a CV and an offer' feature that provides keyword scores, identifies missing terms, and drafts cover letters. For community health centers and NGOs, Mosala includes a 'Community messages' module that translates notices into multiple local languages, with a verification step requiring a native speaker to tick 'Checked by a speaker' before copying. The language support is extensive, covering at least 150 languages across all 54 African countries, with a minimum of three languages per nation.

Key Takeaways

  • Mosala is open-source and available on GitHub at OscarMulume/nonchalant-cv.
  • The tool supports voice dictation, lowering the barrier for users who prefer speaking over typing.
  • It handles face-to-face conversations between two people who share no common language, functioning on one or two phones.
  • The project demonstrates the viability of using small models (e.g., qwen2.5:3b) for specialized, low-latency tasks.

The Bottom Line

Mosala is a masterclass in solving real-world problems with minimal infrastructure. It proves that you don't need a massive cloud budget to make a tangible impact on the global workforce.