People Curated

ML Engineer

📍 Remote

💼 contract

💰 ZAR 120,000 - 150,000 / year

Machine Learning Engineer (Document Intelligence & Fintech) Role Overview We are seeking an entrepreneurial Machine Learning Engineer with production experience to spearhead the development of our unified data management and intelligent document processing pipeline. This role focuses on deploying production-grade computer vision, natural language processing, and rule-based statistical models into a single application engine: Spatial Document Intelligence (Computer Vision): Utilizing OCR and layout-aware deep learning to convert highly unstructured, multi-column, or scanned layout documents into clean JSON schemas while maintaining absolute text grounding and pixel layout coordinates. Contextual ML Processing & Entity Extraction (NLP): Developing and tuning NLP models to automatically parse, classify, and categorize specific documents. Statistical Allocation Engine: Implementing statistical analysis and heuristic-driven validation models Key Responsibilities 1. Model Development & Vision Pipelines (Computer Vision & OCR) Design, tune, and deploy computer vision and image preprocessing pipelines to handle highly variable, scanned, or low-quality document layouts without losing text grounding. Optimize ML models for production, focusing on minimizing inference latency (targeting sub-300ms) when integrated with backend REST APIs. Implement asynchronous task queues to decouple heavy ML inference and OCR processing from the HTTP request/response layer. 2. NLP, Tokenization & Feature Engineering Train, calibrate, and deploy NLP models (ranging from statistical text classifiers like TF-IDF + calibrated regressions to LLM orchestration/RAG) to extract context-specific data from text streams. Build custom tokenizers and preprocessing pipelines to normalize messy, transaction formats, inconsistent merchant strings, and text formats (e.g., SMS alerts). Address model weaknesses systematically through targeted data collection, data augmentation, and dataset refinement to drastically improve precision/recall metrics. 3. Statistical Anomaly Detection & Logic Rules Engineer statistical risk-scoring and fraud-detection systems using outlier detection methods (such as Z-score analysis, round-number bias, and duplicate pattern recognition). Develop the mathematical logic layer for a zero-based asset/deficit allocation engine that checks parsed data streams against strict programmatic constraints. 4. MLOps, CI/CD, & Evaluation Maintain an enterprise-grade CI/CD pipeline with strict testing frameworks (e.g., pytest, GitHub Actions) to catch production bugs using local mock environments. Implement live monitoring endpoints to track per-class precision/recall, data drift, and confusion matrices to drive iterative model improvements.