Malachi Erskine

Applied machine learning practitioner with experience developing computer vision systems used in production. Strong background in data curation, handling subjective labeling challenges, and improving real-world model performance. Combines hands-on ML development (TensorFlow, AWS) with client-facing experience and a deep understanding of user needs.

Image Quality Classification System (Production) 

- Built a computer vision model to evaluate whether images meet defined quality requirements. 

- Integrated into a production workflow processing thousands of images annually. 

- Designed labeling and evaluation strategies to account for human variance in subjective image quality judgments.

Posture / Shoulder Alignment Detection System (Production)

- Developed an end-to-end computer vision pipeline to assess human shoulder alignment from images. 

- Led data collection and curation, building a high-quality dataset tailored to alignment prediction.

- Improved user experience by providing actionable feedback on rejected images, reducing ambiguity and resubmissions., Built full pipeline including preprocessing, inference, and output scoring.

Image Distortion Detection Model (Pre-Sales Demo)

- Architected a model to detect and quantify image stretching/distortion for a high-stakes enterprise sales demo.

- Demonstrated rapid prototyping of ML solutions aligned with business and sales objectives., 

Product Recommendation Tool (Exploration Phase) 

- Conducted early-stage customer discovery to identify pain points in product selection workflows.

- Interviewed users and validated problem areas prior to development.

- Explored potential ML-driven approaches for personalized recommendations.