Applied Machine Learning | Model Reliability & Evaluation | Computer Vision

Give me a messy dataset, a model behaving strangely, or a result that does not quite add up, and I will keep digging until I understand why.

I am particularly interested in applied ML problems where the interesting work starts underneath the headline metric: model behaviour, failure analysis, reliability, experimental design and figuring out why something is behaving the way it is.

I am currently completing an MSc in Computer Science and am based in Israel.

I am building a Python/PyTorch framework for investigating what happens to image classifiers as their inputs move away from ideal conditions.

The framework evaluates models under controlled blur, Gaussian noise and low-light degradation, saving prediction-level evidence so that accuracy, confidence and calibration can be analysed together.

The work currently includes:

• CNN, ResNet18 and MobileNetV2 model support
• transfer learning and full fine-tuning with pretrained ImageNet weights
• controlled GTSRB and MNIST experiments
• track-aware data splitting to prevent validation leakage
• prediction-level failure and confidence analysis
• ECE, confidence-accuracy gap, HCER and failure-detection diagnostics
• reproducible configuration, checkpoint and experiment metadata
• automated testing and experiment safeguards
• trust-rule attribution and reliability analysis across degradation severity

One of the things I am most interested in is where aggregate metrics hide important behaviour at prediction level, particularly when a model becomes more confident in predictions it is still getting wrong.

Online Shopper Purchase Prediction
Built and evaluated models for purchase-intent prediction using scikit-learn, including feature engineering, imbalanced classification and comparison of SVM, Random Forest, HistGradientBoosting and MLP models.

REST API Development
Built an authenticated REST API using Node.js and MongoDB, containerised with Docker and deployed to Google Cloud.

Data & Software Development
Experience across relational databases and SQL, Java and object-oriented development, network/PCAP analysis, Git/GitHub, automated testing and collaborative Agile projects.

ML & Data: Python, PyTorch, scikit-learn, pandas, NumPy, torchvision
Software & Data: Java, JavaScript, SQL, REST APIs
Tools: Git/GitHub, pytest, Docker, Jupyter

I am looking for an ML or ML-adjacent engineering role where I can bring curiosity and grit, dive deep into difficult technical problems, and help turn them into something that actually works.

I am especially interested in computer vision, model evaluation, robotics and autonomous systems, and data-heavy scientific applications.

📍 Israel