Matome Mbowene
Software Engineer — Backend, Applied AI & Data Systems
Johannesburg, South Africa (open to relocation) matomepontso@gmail.com github.com/MatomeMb linkedin.com/in/matomembowene matomembowene.co.za
Profile
Software engineer building production systems that hold up in deployment: validation-first OCR document automation, grounded retrieval (RAG) assistants, and backend platforms anchored in database-enforced correctness. BSc Computer Science & Computer Engineering (University of Cape Town). I care about correctness, auditability and systems that fail safely — and I document what I build so other engineers can verify it.
Verified Outcomes
- Eliminated silent OCR character-swap errors in production ledgers via four-layer validation gates at the API boundary (format, geometry, arithmetic, confidence).
- Reduced the average time required to resolve tutor scheduling conflicts by approximately 35% compared with the previous manual allocation process (UCT Science Learning Centre).
- 89.33% reproducible test accuracy on FashionMNIST with a fully seed-locked PyTorch pipeline and JSON metrics artefacts.
- Zero-hallucination retrieval posture: cosine-gated FAISS retrieval with explicit refusal paths and per-answer citations.
Experience
Fairflow — Co-Founder & Software Engineer
Jan 2026 – Apr 2026- Design validation-first backend services in Python and TypeScript; schema validation executes before business logic at every API boundary.
- Built the OCR document pipeline: OpenCV normalisation (deskew, Otsu binarisation), coordinate segmentation, and confidence-gated ingestion with coordinate-level audit trails.
- Built the grounded RAG assistant: deterministic FAISS indices, pinned embedding models, conservative retrieval gating, explicit refusals.
- Locked environments to pinned lockfiles and instituted structured telemetry for audit-grade operations.
Science Learning Centre, University of Cape Town — Systems Developer
Feb 2025 – Nov 2025- Built the MyAdvisor scheduling system: relational invariants (unique constraints, foreign keys) plus an allocation scoring routine — reduced the average time required to resolve tutor scheduling conflicts by approximately 35% compared with the previous manual allocation process.
- Automated administrative imports with per-row validation, removing manual data-entry error classes.
Education
University of Cape Town
2020 – 2025BSc Computer Science & Computer Engineering. Algorithms, operating systems, computer architecture, embedded systems, networks, machine learning.
Programmes
- Google Cloud Career Launchpad — Cloud Engineering track (2024)
- Dell Young Leaders — leadership development programme (2021 – 2025)
- EY Uncovered — technology consulting insight programme (2025)
Skills
Languages: Python · Java · TypeScript · SQL · C/C++
Backend: Spring Boot · Node.js · FastAPI · REST API design · PostgreSQL · MySQL
AI: PyTorch · FAISS · sentence embeddings · RAG pipelines · OpenCV · LLM APIs
Cloud & DevOps: Google Cloud · Docker · GitHub Actions · CI/CD · Linux
Data: ETL pipelines · Pandas · schema design · validation engineering
Interactive version with role-targeted profiles: matomembowene.co.za/#/resume