Executive Summary

The candidate is a highly capable senior computer vision engineer with a decade of experience across 3D reconstruction, real-time deep learning inference, and graphics engineering. They matches the majority of the job's core requirements — Python, PyTorch/TensorFlow, OpenCV, depth estimation, object detection, inference optimization, and 3D reconstruction — at a level that exceeds mid-level expectations. Their independent 3Vial project is a strong indicator of engineering depth and self-direction. The primary gaps are an absence of explicit SLAM/visual odometry experience and no direct reference to PCL or Open3D, though both are plausibly acquirable given their background. They are recommended as a FIT, with interview validation focused on the SLAM and point cloud tooling gaps before a final offer decision.

Top Strengths

✓Deep, multi-year hands-on experience in 3D vision, reconstruction, and real-time vision pipelines
✓Strong production ML deployment expertise (TensorRT, ONNX, Triton, GPU optimization)
✓Rare combination of computer vision, deep learning, and low-level graphics engineering skills
✓Demonstrated ability to lead and deliver independently — 3Vial project and CV Lead role at Doob
✓Solid academic foundation in Applied Mathematics (MSc) supporting algorithmic rigor

Key Concerns

!Explicit SLAM / Visual Odometry experience not demonstrated — a stated required skill for this role
!No GitHub or code sample provided, limiting direct code quality verification

Culture Fit

75%

Growth Potential

High

Salary Estimate

$80k–$105k (Serbia-based, senior-level profile likely targets upper mid-range for EU remote)

Assessment Reasoning

The candidate meets or exceeds requirements across the majority of the role's stated skills: 3D computer vision, PyTorch/TensorFlow, depth estimation (via photogrammetry), object detection, pose estimation, real-time processing, OpenCV, Python, production inference optimization, and geometry/mesh processing. Their experience level (10 years, including a CV Lead role) is above mid-level, which is a positive signal for quality. Two stated required skills — SLAM/pose estimation and explicit PCL/Open3D usage — are not confirmed on their resume, preventing a higher confidence score. However, their photogrammetry-based 3D reconstruction leadership at Doob and pose estimation work at Continental suggest strong adjacency to SLAM, and their geometry processing work is consistent with point cloud familiarity. The overall skill match exceeds 80% of stated requirements, experience aligns with the role, and no red flags are present — resulting in a FIT decision with a recommendation to probe the SLAM gap in interview.

Interview Focus Areas

SLAM and pose estimation: probe depth of knowledge and any unreported hands-on experiencePoint cloud processing pipelines: assess familiarity with PCL/Open3D specificallySystem design: walk through a real-time 3D vision pipeline design for a production use caseInference optimization: discuss TensorRT/Triton deployment strategies and trade-offs at edge vs. cloud

Experience Overview

10y total · 7y relevant

The candidate is a highly experienced computer vision and graphics engineer whose background directly maps to most of the role's requirements. Their work across 3D reconstruction, real-time inference pipelines, deep learning model development, and production deployment is a strong match. The primary gap is an absence of explicit SLAM/pose estimation experience and no direct mention of PCL or Open3D libraries, though their photogrammetry and geometry processing background suggests strong adjacency.

Matching Skills

Python (expert)PyTorchTensorFlowOpenCVObject DetectionImage SegmentationPose EstimationReal-time Processing3D Computer VisionPhotogrammetry / 3D ReconstructionDepth Estimation (adjacent via photogrammetry)Point Cloud Processing (adjacent via geometry/mesh processing)ONNX Runtime / TensorRT (inference optimization)GPU Computing / CUDA

Skills to Verify

Explicit SLAM / Visual Odometry experienceDirect PCL / Open3D library usage not mentionedPoint cloud processing not explicitly called out (implied via photogrammetry and geometry work)

Candidate information is anonymized. Personal details are hidden for fair evaluation.