I'm a 19-year-old self-taught developer building Mothusi — an AI voice and chat study assistant helping BGCSE Computer Studies students in Botswana learn and test themselves, for free — and Tsela, an offline AI admissions assistant built for the Africa Deep Tech Challenge 2026.
My name is Amantle Mpaekae, also known as Mulax Prime. I'm a self-taught developer from Mogoditshane, Botswana, working toward becoming an AI Engineer.
I scored a D in Computer Studies for BGCSE. That same subject is what I now build software for. Mothusi exists because I know what it feels like to be stuck on a topic with no one to explain it simply — so I built the explainer I needed. Tsela exists because I went through the university admissions process myself with no reliable source for points requirements and deadlines — so I built that too, offline, for the next student going through the same thing.
I learn by shipping, not by studying theory first. Every version of Mothusi and every constraint on Tsela taught me something no course could — debugging real crashes, fixing real architecture problems, demoing to real teachers, and getting real feedback.
Terminal app, voice only. Demoed at Mogoditshane Senior Secondary — failed first due to a missing API key, fixed same day. First GitHub release.
Full GUI, animated orb, student registration, 7-question quiz with scoring. Demoed to a senior teacher who gave real feedback.
Rebuilt from scratch. Google AI-style interface, login state machine, typing and voice input, natural AriaNeural voice.
5 bugs fixed, quiz save logic rewired, syllabus updated with 2024 BGCSE examiner report weak topics. Current version.
Desktop applications with tkinter, state machines, threading, and clean modular architecture. Built and shipped, not just learned.
Connecting applications to Groq AI for fast, context-aware responses, and running quantized local LLMs offline via llama.cpp for zero-connectivity environments.
Speech recognition for input and natural text-to-speech for output, built into real applications students actually use.
An AI voice and chat study assistant for BGCSE Computer Studies students in Botswana. Students type or speak a question and get an answer read aloud in a natural voice. Includes a full quiz mode — 7 questions, 10 marks, two attempts per question, and automatic score saving. Built with Python, tkinter, Groq AI, edge-tts and SpeechRecognition. Demoed live at Mogoditshane Senior Secondary School.
An offline, on-device AI assistant answering university admissions questions for six Botswana institutions. Runs a quantized Qwen2.5-1.5B model via llama.cpp on an 8GB no-GPU laptop, with a deterministic lookup that bypasses the model entirely for points and requirement questions — preventing hallucination on the numbers that matter most. Built and submitted solo for the Africa Deep Tech Challenge 2026.
Discovered an interest in technology and problem solving — started exploring how computers actually work.
Completed BGCSE. Started self-teaching Python during school and began building real projects while still a student.
Building Mothusi and Tsela solo. Working toward AI Engineering through real projects rather than courses — learning by shipping and fixing bugs in production.
Fixed five production bugs, rewired the quiz save logic, updated the syllabus with 2024 examiner report weak topics, and demoed to a Computer Studies teacher.
Built and submitted an offline AI admissions assistant solo, including official performance profiling, on a budget-tier laptop with no GPU.
Interested in Mothusi or Tsela, want to collaborate, or just want to connect with another Botswana developer? Reach out.
Currently focused on Mothusi and Tsela.
I'm not taking freelance bot or web development work right now — all my time goes into building these projects and learning the next stage of the roadmap above.
If you want to follow the projects, check the GitHub repos or reach out by email.