Matome Mbowene
Software & AI Engineer, production OCR / document automation, retrieval (RAG), backend systems
100%
OCR field-mapping accuracy
89.33%
FashionMNIST test accuracy
35%
Scheduling efficiency gain
Summary
I build production-focused AI features that ship: document automation, retrieval pipelines, and services, measured, validated, and designed for reliability. I’m comfortable across the stack and I optimize for correctness, maintainability, and clear delivery.
Selected outcomes (public-safe)
- 100% OCR field‑mapping accuracy on a defined document set using validation-first design.
- 89.33% FashionMNIST model accuracy (public demo context).
- 35% scheduling efficiency improvement (public-safe summary).
Experience (public-safe highlights)
Computer vision & document automation
Recent
- Built a validation-first OCR pipeline with confidence scoring and auditability.
- Designed layered checks and fallbacks to reduce silent extraction errors.
Retrieval systems (RAG)
Recent
- Developed a RAG assistant pipeline with reproducible indexing and conservative guardrails.
- Improved “proof” signals by linking outputs back to sources where possible.
Embedded / edge fundamentals
Recent
- Contributed to latency-aware, reliability-first pipeline design (public-safe summary).
Timeline (high level)
Production OCR systems
Validation-first extraction, auditability, and reliability guardrails.
Retrieval systems (RAG)
Deterministic indexing, evidence-first responses, and safe fallbacks.
Backend + systems work
APIs, data modeling, and systems-level performance considerations.
Some work is under NDA; I can discuss engineering approach and outcomes at a high level.