Fennec Signal Independent deep-tech consulting

About Fennec Signal

Boaz Schwartz, Ph.D. — independent AI / ML systems architect and algorithm consultant

An AI and algorithm specialist with a background spanning statistical signal processing, machine learning, deep learning, system architecture and high-performance AI systems.

Ph.D. in statistical estimation and learning algorithms. Professional work across LiDAR and autonomous systems, audio and microphone arrays, biological and environmental sensing, predictive maintenance, medical applications, embedded systems, cloud infrastructure and AI.

Independent consultant and architect, often embedded in R&D teams for long engagements.
From early algorithmic exploration and feasibility studies through architecture, simulation, prototyping, performance analysis and production-oriented development.
Starting from the mathematics and the signals, understanding the complete system, then building efficient implementations tested against real data and real hardware.

AI / ML Systems Architecture & LLM Algorithms

I build algorithms and AI systems where deep mathematical understanding meets practical, efficient implementation.

Current work focuses on LLM inference and serving systems, built upon a long background in statistical algorithms, machine learning and system architecture.

LLM inference & serving
GPU · memory · network co-analysis
Coding-agent driven R&D
Statistical algorithms

Three layers of an AI system: the hyperscaler datacenter, the server board, and the low-level implementation in software, firmware and hardware

AI-Native Algorithmic R&D

Coding agents and AI-assisted environments as engineering tools, not as a faster typewriter.

Modern coding agents close the loop between mathematical reasoning, implementation, simulation, experimentation, analysis and validation. The loop that used to take a sprint now takes an afternoon, so the question becomes which larger problem to put inside it.

The point is not to generate code faster. It is to tackle substantially larger algorithmic and system problems with much shorter development cycles: more hypotheses tried, more of the system simulated, more of the analysis automated and checked.

Agentic engineering workflows around difficult algorithmic problems
Automated experimentation, sweeps and regression checks
Simulation and analysis tooling built alongside the algorithm
Research-to-code: from derivation to a tested implementation
AI-assisted debugging and validation against real data and real hardware

The AI-driven development life cycle as a loop: plan and design, AI coding, intelligent testing and autonomous deployment, branching to a new path forward when the performance is not enough — The AI-driven development life cycle: plan and design, AI coding, intelligent testing and autonomous deployment — and round again, by a new path, when the performance is not yet enough.

Example Projects

Medical imaging, optical, acoustic and wearable sensing.

Cerebral vessels reconstructed in three dimensions from a CT angiography, beside the viewer the segmentation was tagged in

Brain Blood Vessels for Stroke — CT Analysis

Co-registration of CTA, NCCT and 4D CTP series of the same head onto one grid
Vessel segmentation and vascular characterization of the cerebral arteries
Perfusion maps — TTP, CBF, CBV, MTT, Tmax — from the time-attenuation curves, validated against clinical reference data

Silicon photonics chip with light paths across the die

Optical Signal Processing

Lidar signal processing: noise reduction, reflectivity estimation, blooming reduction
Optical finger sensor: spectroscopy-based unintrusive alcohol detection
Optical computer: LLM-optimized photonic AI using silicon photonics

Acoustic laboratory with a microphone array under measurement

Acoustics & Microphones

Microphone arrays with directional analysis and beamforming
Wearable audio technologies with real-time noise reduction
Room acoustics measurement and modeling for studio/lab environments

Smart safety helmet with IoT module and embedded sensors

Mobile & Wearable Sensing

Context-aware mobile environment classification (car, indoors, and more)
Sensor fusion for bio-acoustics and behavior insights
Chewing activity sensing using AirPods microphones

Client Engagements

Long-term engagements, embedded in the client's R&D team.

Resolight — photonic processing, scale everywhere

AI Datacenter Network Simulation

Simulation of LLM training and inference workloads on GPU clusters — parallelism, collective communication and fabric traffic — to size the networks that Resolight's all-optical switching is built for.

Lidar System Architecture & Signal Processing

System architect for automotive LiDAR: physical-system behaviour translated into algorithmic and system requirements; noise reduction, reflectivity estimation, blooming reduction.

Wearable Audio Technologies

Wearable — noise reduction, microphone-array processing, real-time implementation.

Industrial stamping machine with strain sensors

Stamping Process Monitoring

Strain sensors & HMM analysis for fault detection; predictive maintenance from statistical modeling of real production data.

A smartphone opened to its sensors, each wired to an environmental reading

Phone-Based Environmental Sensing

Mobile Physics turns the phone's own sensors into a hyper-local environment monitor — temperature, humidity, air quality, light, UV and noise — with no extra hardware. Algorithms for context-aware sensing and sensor fusion on the device.

Academic Work

Studies, publications and teaching.

Publications & Patent

Papers on statistical speech and audio algorithms — recursive expectation-maximization, blind source separation, concurrent speaker tracking, speech dereverberation and hearing-aid processing — in speech and audio conferences and journals, among them Interspeech and the EURASIP Journal on Audio, Speech, and Music Processing.

Patent: Method and System for Acoustic Source Enhancement using Acoustic Sensor Array US 10,334,390 B2 · issued June 2019

Teaching

Lecturing and supervision, inside the university and outside it.

Lecturer and supervisor, Bar-Ilan University (2019–2020): signal processing and machine learning courses, and supervision of research and algorithm-development projects that connect statistical theory with working engineering.
Lecturer in computer science, ultra-Orthodox student group (2018–2020): three years teaching computer science to a Haredi group, from programming foundations to algorithmic thinking.

Courses

Statistical Signal Processing
Machine Learning
Programming Foundations
Algorithmic Thinking
Research & Project Supervision

Special Projects & Unique Cases

Industrial & biological monitoring systems.

Carrots flying out of an optical sorting drum, sorted by air pushers

🥕 Industrial Carrot Sorting

Real-time quality monitoring in production lines using computer vision and signal analysis to prevent waste and improve efficiency.

🐣 Hatchery Monitoring & Control

Optical monitoring for poultry hatcheries, enabling precise biological development tracking and optimal environmental control.

Let’s Build Your Next AI or Sensing System

From the mathematics and a feasibility study to simulation, architecture and a deployment-ready implementation, Fennec Signal helps R&D teams take on the larger problem.

📞 +972-50-876-2411 ✉️ boaz@fennecsig.com in LinkedIn 📍 Mazkeret Batya, Israel