Deep Learning & Model Training
PyTorch, foundation model fine-tuning, contrastive/triplet learning, embeddings, model evaluation.
Korin Lifshits · Senior Computer Vision & AI Engineer
Engineering Robust Perception for the Physical World.
Senior AI & Computer Vision Engineer specializing in 3D vision, deep learning, multi-object tracking, ReID, calibration, and real-world perception systems.
About
Applied research discipline with production engineering judgment.
KoriVision is the personal professional website of Korin Lifshits, a senior engineer working across classical geometry, 3D reconstruction, tracking, ReID, deep learning, model training, and production-oriented perception systems.
The work centers on perception systems that must operate outside clean benchmarks: constrained sensors, changing domains, ambiguous identities, timing budgets, imperfect calibration, and failure modes that only appear in the physical world.
Expertise
Core technical pillars
A practical blend of learning-based methods, geometric reasoning, and system-level reliability work for real-world perception products.
3D Vision & Geometry
Calibration, SfM, SLAM, point clouds, ICP, multi-view reconstruction.
Tracking & ReID
Kalman filters, multi-object tracking, Hungarian matching, cross-camera association, identity preservation.
Real-World Systems
Latency, robustness, sensor constraints, data drift, reliability, failure analysis.
Abstracted Systems Experience
Public-safe engineering themes
Selected experience is intentionally described at a high level, without confidential employer details, customer names, proprietary architectures, internal datasets, or sensitive operational specifics.
Distributed perception systems
Architecting perception pipelines that coordinate multiple sensing, inference, and decision stages under practical reliability constraints.
Medical imaging and precision tracking
Developing careful vision workflows for measurement-oriented imaging contexts where repeatability and validation discipline matter.
Multi-camera 3D reconstruction
Combining calibration, multi-view geometry, point-cloud reasoning, and registration to recover structure from constrained viewpoints.
Real-time video analytics
Balancing accuracy, latency, throughput, and observability for video systems that need dependable behavior over long-running streams.
Domain adaptation for ReID
Improving identity representations across camera conditions, visual domains, and deployment environments while monitoring drift and failure cases.
Research Notes / Digital Garden
Ongoing theoretical exploration
Unfinished notes and working ideas around mathematical structure, perception geometry, and diagnostics. These are framed as intellectual exploration, not product claims.
Ecosystem
Related public-facing work
KoriVision
Personal professional portfolio for senior computer vision and AI engineering work.
AIZYON
AI/CV knowledge hub for structured notes, concepts, and practical learning material.
PersenseAI
Visual agents and perception diagnostics lab focused on interpreting and testing perception behavior.
AIVMeta
AI media and short educational content around applied computer vision and AI systems.