Hi, my name is Hanoch Kremer.

Team leader — Computer Vision | AI | ML | LLM | Deep Learning | Signal Processing | Mentor
AI CV Team Leader at Rail Vision Ltd

Bio

Applied scientist with 20+ years of end-to-end research and product delivery across computer vision, NLP, ASR, and signal processing.

I specialise in translating state-of-the-art research into production-grade AI products at scale, delivering value across global enterprises (Huawei, MSD) and early-stage startups (Innoviz, Autotalks, Xsight), from concept to customer use.

I have proven deployment breadth across edge (SIMD/mobile, ASIC, car-kit ASR), cloud, and on-premises environments. As an experienced research lead, I mentor teams, align R&D roadmaps with business objectives, and collaborate with cross-functional stakeholders.

I was part of the founding teams at two automotive technology startups; both achieved successful exits (acquisition and IPO). I hold nine commercially successful patents spanning deep learning, visual perception, speech enhancement, and signal processing.

As a team leader, tech lead, IC, or mentor in cutting-edge teams, I collaborate effectively with cross-functional teams and stakeholders. I mentor in the (AI) Adelson entrepreneurship program and serve on a graduates’ examination committee. I served in an elite technology unit in the IDF. I am an EU resident.

My proficiency

AI / NLP / text

• GPT/LLM-based semantic summarization of scenes
• Generative AI: multi-modal understanding and retrieval with LLMs
• Semantic search and retrieval; fine-tuning LLMs and parameter-efficient fine-tuning (PeFT)
• Hands-on work with transformer-based models

Computer vision, deep learning, and ML

• Perception: detection, classification, and tracking for 2D, 2.5D, 3D, and thermal (LWIR) modalities
• Re-identification and novelty detection
• Remote sensing, change detection, and multi-sensor fusion
• Stereo vision, camera anatomy, calibration, and image processing
• ML, statistical modeling, and optimization

Speech recognition and processing

• ASR in non-stationary, harsh environments: noise compensation and speaker/model adaptation with sparse data (e.g. car kit)
• Adaptive voice activity detection (VAD) for harsh environments
• Speech enhancement for vehicle environments; acoustic echo cancellation

Signal processing

• RADAR (FMCW), noise cancellation, adaptive filtering and interference cancellation

Toolbox

PyTorch, TensorFlow, Keras, Python, Caffe, OpenCV, Weka, C/C++, MATLAB, parallel programming, Assembler, Linux (user), LibSVM, SQL, vector databases, Git, AWS, and MLOps. Comfortable across software and hardware-oriented stacks: OOD, ASIC/VLSI, IPP, and VLIW/SIMD parallel programming. Agile/SCRUM.

In my free time I swim with a Masters group several times a week, ski when I can, and play drums in a band. I am married with two wonderful teenagers.