QA Annotation Analyst

The 4M story is likely one you haven't heard before: we are on a mission to reveal the world below us — to do for the world below ground what Google Maps did for the world above. By leveraging cutting-edge computer vision, we map subsurface infrastructure to make reliable, real-time utility data accessible to the construction industry, completely transforming a traditional industry. We're a growing startup with 100 employees based in Tel Aviv, Israel, and Austin, Texas. Our algorithms learn from data — and the quality of that data is what determines the quality of the maps we produce. Annotation quality is not a back-office function; it is the foundation the entire mapping engine is built on. Every object our models learn to detect starts with a human labeling it correctly, and every labeling error compounds downstream. This role sits at that foundation. The Opportunity We are looking for a QA Annotation Analyst to join our core Algorithm team and own the quality of the annotated data that trains our computer vision models. You will be the person who ensures that what goes into our models is accurate, consistent, and trustworthy — reviewing labeled subsurface and street-level imagery, catching errors before they reach training, and continuously raising the quality bar of our datasets. This is a hands-on role for someone with a sharp eye for visual detail, strong judgment about edge cases, and the discipline to apply annotation standards consistently across large volumes of data. You'll work closely with the algorithm engineers who consume the data and with the teams and vendors who produce it. What You'll Do • Review annotated computer-vision datasets (street-level panoramic and geospatial imagery) for accuracy, completeness, and consistency against the annotation guidelines. • Identify, document, and categorize labeling errors to raise systematic issues, ambiguous cases, and gaps in the guidelines back to the algorithm team. • Serve as the quality gate between annotation production and model training: decide what meets the bar and what needs rework. • Help refine and maintain annotation guidelines as new object types and edge cases emerge. • Track and report quality metrics (error rates, class-level accuracy, consistency) so the team has a clear, ongoing picture of dataset health. • Provide clear, actionable feedback to internal annotators and external annotation vendors to drive quality up over time.

Requirements: Required Qualifications • 1–3 years of experience in data annotation, QA, image review, or a closely related data-quality role. • Sharp visual attention to detail — able to spot subtle labeling errors and inconsistencies across large volumes of imagery without losing focus. • Strong, consistent judgment: applies standards uniformly and reasons carefully about ambiguous or edge-case decisions. • Comfortable working with annotation and labeling tools, and with reviewing structured visual data at scale. • Organized and reliable — keeps clear records, tracks issues systematically, and follows through. • Good communication skills — can explain quality issues clearly and give constructive feedback to annotators and vendors. • Full-time or minimum 80% availability. Preferred Qualifications • Experience annotating or reviewing computer-vision datasets (bounding boxes, segmentation, keypoints). • Familiarity with GIS or geospatial data, maps, or infrastructure imagery. • Exposure to the ML/CV data lifecycle — understanding how annotation quality affects model training and performance. • Experience working with or overseeing external annotation vendors. • Basic scripting or spreadsheet skills for tracking and analyzing quality metrics. Why Join 4M If you take pride in getting the details right and want that precision to directly shape a product redefining an industry, this is that role. • Your work is the foundation of the product — the quality you enforce is what our models learn from. • You'll sit inside the algorithm team, close to the engineers who depend on your judgment — not siloed away from the work that matters. • Real ownership of dataset quality, with the autonomy to shape how QA is done as the team scales. • Pre-scale startup (~100 people): your influence on quality standards and process is real and lasting. • A domain undergoing a once-in-a-generation transformation — and a team that values rigor and gets excited about hard problems.