My Resume

Work Experience

Microsoft - Computer Vision Data Scientist PhD Intern

Nov/2025 - Present | WSSI

Developing a production-grade image super resolution and restoration system as part of a cross-functional team.
Conducting comprehensive literature reviews and benchmarking CNN-based and Transformer-based architectures on internal datasets.
Training, fine-tuning, and optimizing models, including knowledge distillation techniques to compress high-performing architectures for deployment.
Designing and executing rigorous evaluation pipelines to assess model quality across perceptual and fidelity metrics.

Elop - Computer Vision & AI Algorithm Developer

Feb/2023 - Oct/2025

Performed a diverse range of Computer Vision and Vision-Languege tasks and developed a deep understanding of various state-of-the-art (SOTA) models and architectures.
Received project requirements and system constraints, then designed, optimized, and integrated custom algorithmic pipelines into larger systems.
Conducted extensive literature reviews—evaluating over 70 research papers annually on customized data with numerical performance metrics.

Scene Understanding and Threat Recognition System

Built an end-to-end pipeline that ingests multi-object tracking outputs and leverages Video-LLMs for complex-scene understanding and per-object threat scoring.
Led a 3-person project; authored and executed a one-year roadmap (versioning, scope per release, task allocation) to deliver the project.
Benchmarked multimodal models from 3B–78B (qwen2.5-VL, LLaVA-OV, etc.) on multiple datasets, comparing accuracy and runtime/latency to select deployment candidates.

Multi-Object (Multi-Sensor) Tracking

Developed a tracking solution that increased precision from 72% (naive off-the-shelf YOLO+DeepSORT pipeline) to 93% using a customized pipeline.
Integrated an ensemble of video classification models (MViT, TimeSFormer, etc.) focused on reducing false alarm rates.
Implemented a three-sensor tracking system (RGB, SWIR, MWIR) to ensure robust performance across different imaging modalities.

Video Description: Visualization of simple basic off-the-shelf multi-object tracking in RGB (Vis) sensor without the full pipeline (Due to IP and confidentiality).

Zero-Shot Keypoint Tracking

Explored zero-shot tracking for keypoints in infrared videos characterized by extreme noise conditions.
Utilized simulative data generated with UE5 specifically for testing under high interference scenarios.
Rigorously evaluated SOTA models to determine limitations and ensure their suitability under adverse conditions.

Video Description: Visualization of tracking keypoints on a passenger airplane from standart RGB (no noise) video (Due to IP and confidentiality).

Video Restoration and Super Resolution

Investigated and compared multiple SOTA models, including a self-designed architecture and a self-trained model.
Applied knowledge distillation from a FeMaSR teacher model to optimize the chosen architecture.
Achieved performance improvements of 5–20% over classical methods, as measured by metrics such as MTF, SSIM, and ESF.

Video Description: Demonstration of video restoration and turbulence mitigation. This output is from the DATUM model (Zhang et al., CVPR 2024), as my own work’s data is confidential.

Video Classification for False Target Filtering

Designed a comprehensive data management infrastructure for experimental model training and evaluation.
Developed tools for dynamic data integration, version management, attribute-based filtering, and complex augmentations (e.g., smart copy-paste and 3D rotations relative to the camera plane).
Conducted hundreds of training experiments across both lightweight and heavy SOTA architectures, rigorously evaluating each component (data version, augmentations, architecture, transfer learning, etc.).
Increased overall accuracy by 22%.

Ben-Gurion University - Teaching Assistant

2020 - 2023

Algorithms and Graph Theory (2023)
Digital Design (2022)
Digital Computers Structure (2022)
Linear Algebra (2020-2021)

Education

Ph.D. in Electrical and Computer Engineering

Ben-Gurion University, 2021 – Present

Lab: InsightLab

GPA: 97

Advisors: Prof. Kobi Cohen (2021–present), Dr. Chaim Baskin (2024–present)

Accelerated Ph.D. program for top-performing students.

Main courses: Deep learning, Sequential learning, Statistical inference and Data Mining, Multivariate statistical data analysis, Game theory.

B.Sc in Computer Engineering

Ben-Gurion University, 2018 – 2022

GPA: 92

Honors: Summa Cum Laude (in the dean's list)

Ranking: Ranked #1 in my class (2018–2022 BGU CE)

Research: Researched computational modeling of protein-DNA/RNA interactions using deep learning models, guided by Dr. Yaron Orenstein.

Publications

HiMu: Hierarchical Multimodal Frame Selection for Long Video Question Answering

Dan Ben-Ami, Gabriele Serussi, Kobi Cohen, Chaim Baskin

Under review, 2026

We introduce HiMu, a training-free hierarchical multimodal frame selection framework for long-video QA that decomposes queries into logic trees with specialized experts, advancing the efficiency-accuracy Pareto front while requiring ~10x fewer FLOPs than agentic systems.

HERBench: A Benchmark for Multi-Evidence Integration in Video Question Answering

Dan Ben-Ami, Gabriele Serussi, Kobi Cohen, Chaim Baskin

Accepted to CVPR, 2026

We present HERBench, a VideoQA benchmark for multi-evidence integration across time, requiring aggregating multiple non-overlapping evidential cues. Evaluating 13 state-of-the-art Video-LLMs reveals pervasive failures (31-42% accuracy vs. 20% random baseline), disentangled into retrieval and fusion deficits.

A Stable Polygamy Approach to Spectrum Access with Channel Reuse

Dan Ben Ami, Kobi Cohen

In preprint, 2024

I introduced the "Stable Polygamy Problem" (SPP) for spectrum access with channel reuse, developed efficient algorithms including RP&R, and proved their performance in specific interference regimes with strong simulation results.

Client Selection for Generalization in Accelerated Federated Learning: A Multi-Armed Bandit Approach

Dan Ben Ami, Kobi Cohen, Qing Zhao

IEEE Access, 2025

I developed a multi-armed bandit–based algorithm (BSFL) for federated learning that balances training latency and model generalization, achieving logarithmic regret and outperforming prior methods on synthetic and real datasets.

A Universal System for Boosting Gene Expression in Eukaryotic Cell-Lines

Inbal Vaknin, Or Willinger, Jonathan Mandl, Hadar Heuberger, Dan Ben-Ami, Yi Zeng, Sarah Goldberg, Yaron Orenstein, Roee Amit

Nature Communications, 2024

Designed and trained deep learning models to predict protein expression from DNA sequences, guiding motif selection for cross-species promoter boosting.

Characterizing Regulatory Grammar Rules in S. cerevisiae Using a Library of Conserved and Unknown Motifs

Inbal Vaknin, Hadar Heuberger, Dan Ben-Ami, Or Wilinger, Yi Zeng, Leon Anavy, Zohar Yakhini, Sarah Goldberg, Yaron Orenstein, Roee Amit

Cerevisiae, 2023

Developed convolutional neural network models to predict gene expression from DNA sequences, enabling validation of conserved transcriptional grammar rules.

Dan Ben Ami

About Me