Netanel Eliav's Website
01/16/2026 10:47

Projects

SightBit

SaaS (TRL-9)
Public
  • Advanced computer vision platform for real-time drowning prediction, human and vessel detection, and flood-risk alerts.
  • Utilizes standard CCTV feeds, eliminating the need for specialized sensors, edge boxes, or site-specific tuning.
  • Features a panoptic-segmentation model with a proprietary augmentation pipeline that enhances small-object recall by 18 percentage points.

AdaptiveBridge

Production-ready (TRL-8)
Public • Open-Source
  • Python library for adaptive modeling that predicts and fills missing features during inference, ensuring ML models remain robust in real-world applications.
  • Offers configurable thresholds for feature importance, correlation, and accuracy; supports custom accuracy logic and automated distribution handling.
  • The drop-in estimator requires fitting only once, allowing seamless integration with existing pipelines.
  • Effectively manages feature dependencies—mandatory, deviation, leveled—to optimize model performance.
  • Comprehensive documentation, unit tests, and CI included, released on PyPI under a permissive license.

MHIL

Concept & Early Prototyping (TRL-5)
Semi-Stealth • Non‑Profit
  • Machine Human Intelligence Lab (MHIL) is a non-profit institute focused on establishing benchmarks and evaluation protocols for genuine machine intelligence that extends beyond narrow LLM tasks.
  • Designs human-aligned test batteries for agents, tools, and composite systems, addressing memory, planning, robustness, and autonomous problem-solving in uncertain environments.
  • Develops an open evaluation matrix for 'pre-superintelligent' systems, enabling straightforward comparisons across models, vendors, and architectures.
  • Aims for future 'superintelligence-ready' standards, including reproducible scorecards, capability tiers, and safety-critical thresholds for real-world deployment.

MazeByte

SOTA Proof of Concept (PoC)
Semi-Stealth
  • Demonstrating an autonomous, self-maintained AI agent for data automation through fully ETL processes.
  • Implements infant-like feedback loops (reward/pain) and resource-seeking behaviors for adaptive learning.
  • Generates pseudo-code modules for memory management, search optimization, and comprehensive ETL automation.
  • Aims to eliminate manual data engineering by automating the entire data pipeline using the proprietary MBI model.

Jhive

Production (TRL-7)
Public • AIforGood
  • Django-based web application employing OpenAI API for the analysis and detection of antisemitic content.
  • Real-time content classification paired with a detailed reporting interface.
  • Offers RESTful API endpoints for integration with rate limiting.
  • Features secure input validation and thorough sanitization measures to prevent injection attacks.