Pravallika Gaddam

I’m an Amazon-trained operations professional who sustained a 98%+ accuracy rate across 500+ daily data reviews at one of India’s highest-throughput ML operations. I deliver enterprise-grade precision, proactive communication, and zero-handholding execution to remote teams.

In my ML Data Operations role, I raised computer-vision reliability by applying a self-developed 3-step pre-check protocol before every submission. I reduced QA re-review cycles by ~20% across a team of 20+ associates by identifying and documenting 4 recurring mis-labelling patterns and sharing a fix guide with the team lead. I also eliminated decision ambiguity by building structured audit trails on every review using internal tooling.

I delivered zero SLA breaches across the full engagement by building a personal daily workflow dashboard that flagged backlog risk 2 hours before deadline. I contributed reusable documentation templates to standardize the workflow and cut supervisor spot-check time through clearer, traceable records.

3D Annotation Specialist | Freelance — Thron Autonomous Vehicle Project

August 2024 – January 2025 | Remote

• Annotated LiDAR point clouds and 3D mesh data for autonomous vehicle AI training datasets — placing precise 3D bounding boxes around vehicles, pedestrians, cyclists, and road infrastructure across complex urban and highway driving scenarios.Performed semantic segmentation on 3D point cloud data — categorising drivable surfaces, sidewalks, curbs, and static obstacles to support spatial reasoning model training.Reviewed and cleaned raw LiDAR sensor data, identifying and removing noise artifacts from rain, dust, and sensor interference to ensure clean training datasets.Validated simulated 3D environments against real-world point cloud captures, identifying discrepancies and flagging fidelity issues for correction.Performed sensor fusion annotation correlating LiDAR, camera, and radar data streams for multi-modal AI training, ensuring spatial consistency across all sensor inputs.Applied occlusion handling, truncation tagging, keyframe annotation, and panoptic segmentation across 1,500+ daily tasks at peak volume — maintaining 95%+ accuracy and guideline consistency throughout a 6-month engagement.Worked independently across CVAT, Label Studio, Scale AI, Labelbox, SuperAnnotate, and Supervisely with minimal ramp-up on each platform.

Before that, as a B.Tech Mechanical Engineering class representative and department coordinator, I managed inboxes, scheduling, and event coordination for 60+ students and 12+ faculty members. I maintained 100% on-time delivery of departmental events and academic submissions across 8 semesters using a shared calendar system and a 2-week rolling task tracker. I kept communications tight by resolving scheduling conflicts early and drafting 50+ formal announcements and meeting summaries over multiple academic years.