About

JOB DESCRIPTION We operate a commercial vegetable harvesting operation and have cameras installed on our harvesting trailers. We are seeking an experienced computer vision / machine learning engineer for a Phase 1 proof-of-concept (POC) to assess the feasibility of extracting simple, reliable productivity signals from these camera feeds. This phase is explicitly limited in scope and is intended to answer the following questions. Is this feasible in our real-world environment? What is the simplest reliable approach? What level of accuracy is “good enough” for operational decision-making? This is not a full production build. --- PHASE 1 OBJECTIVES The goal of this POC is to determine whether we can achieve the following. 1. Produce a rough but stable estimate of harvesting productivity. Expressed as heads per hour (or similar throughput metric). 2. Identify high-level trailer states, such as: * Actively harvesting * Travelling / repositioning * Stationary / idle 3. Output a basic machine-readable signal (for example JSON, CSV, simple API, or documented interface) suitable for future integration with our crop management system. --- WHAT SUCCESS LOOKS LIKE FOR PHASE 1 At the end of this POC, we expect the following outcomes. * A clear yes / no / maybe on feasibility * A recommended simplest viable technical approach (AI or non-AI) * Example outputs using real sample footage A brief written summary covering: * What works * What doesn’t * Key risks or limitations * Rough effort required for a production system Per-frame accuracy is not required. Stability and usefulness matter more than precision. --- IMPORTANT CONTEXT * Outdoor agricultural environment * Variable lighting, dust, and movement * Cameras and harvesting rigs are mobile, and mobile connectivity is typically patchy with black spots We strongly prefer simple, robust solutions over complex models. If this can be solved without heavy ML, that is a positive outcome. --- SKILLS AND EXPERIENCE REQUIRED * Proven experience in computer vision for real-world or industrial environments * Experience evaluating feasibility and technical trade-offs * Ability to explain technical decisions clearly to non-ML stakeholders Nice to have experience includes: * Agricultural or industrial computer vision * Edge versus cloud inference --- ENGAGEMENT STRUCTURE AND BUDGET (PHASE 1 ONLY) * Hourly engagement * Initial cap of 20–30 hours * Target hourly range of USD $60–$120 per hour, depending on experience This phase is intentionally capped. If Phase 1 is successful, there is a strong likelihood of follow-on work for pilot and production phases. --- HOW TO APPLY Please include the following in your proposal. 1–2 examples of similar computer vision or counting projects. Your high-level approach to evaluating this problem. What you would aim to prove or disprove during a POC. Any immediate risks you see. Generic or copy-paste proposals will be ignored.