A
Computer Vision Model Audit & Optimisation
Adaeze Eze
Freelance · Mid-level · Chicago
About the role
Project scope: Independent technical audit and optimisation of an existing computer vision model used in a manufacturing quality control application. Background: The model has been in production for 14 months. Initial accuracy on the validation set at launch was 94.1%. Current accuracy on production data is estimated at 87.3% based on internal sampling. Root cause is not yet established — candidates include data drift from a change in production line lighting conditions in month nine, label quality degradation in the training dataset, and possible overfitting to seasonal product variations. Deliverables: (1) Written audit report identifying root cause(s) with supporting evidence and analysis. (2) Remediation recommendation with options ranked by expected impact and implementation effort. (3) Implementation of the highest-priority remediation. (4) Post-implementation performance evaluation against baseline. Timeline: Six to eight weeks. Budget: Fixed price. Access: Full access to training datasets, production logs, model weights, and the internal ML engineer who built the original system. The project is well-scoped. We're not looking for someone to redesign the system — we need someone who can diagnose what's wrong and fix it.
Key Deliverables
List the expected deliverables for this project.
- Analyse production logs and sampling data to characterise the accuracy degradation pattern
- Audit the training dataset for label quality and distribution shifts relative to production data
- Produce a written root cause analysis with ranked remediation options
- Implement the agreed remediation and validate performance improvement
- Deliver a final report and conduct a handover call with the internal ML engineer
Requirements
Technical stack needed for this mission.
- Demonstrable experience auditing and improving production computer vision models
- Strong PyTorch for model evaluation, fine-tuning, and debugging
- Experience identifying and addressing data drift and distribution shift
- NumPy and OpenCV for dataset analysis and image preprocessing
- Able to produce clear written technical reports accessible to a mixed technical/non-technical audience
- Manufacturing or industrial inspection domain experience is a plus but not required
Contract Type
Fixed-price
Level
Mid-level
Budget Range
$10,000 – $18,000
Duration
1 – 3 months
AI Expertise
AI & Machine Learning Engineers
Ready to apply for this role?
Create a free talent account in under 2 minutes.
- Apply to verified AI companies
- Get AI-matched job recommendations
- Message hiring managers directly
- Build your public AI talent profile