Selected work

The engagements behind the claims: forecasting, vision, closed-network AI, and operations systems, each described at the exact verb tier of what happened.

A three-year data-science engagement for a Mueller-group industrial manufacturer (US market). Forecasts graded against actuals every month; CEO recommendation on file.

Read the full case study

A full operations transformation for a live-performance company: its whole operation rebuilt off WhatsApp onto one Monday.com and Make.com system, live and in daily production use.

Read the full case study

Every card below uses the highest truthful verb and nothing above it.

Built, delivered, and runs in production

Full operations transformation for a live-performance company

Rebuilt a live-performance company's operations off WhatsApp and a dozen fragmented tools onto one Monday.com and Make.com operating system: around thirty automation scenarios in six workstreams, recurring work that spawns and routes itself, a semi-automated payment chain that rolls attendance, lessons, and supplier invoices into one monthly surface with frozen rate snapshots, and a dedicated control board watching every automation. Two years of operating history migrated. Live and in daily production use.

  • Operations
  • Monday.com
  • Make.com
  • Automation
  • Observability

Architected; assistant built and demonstrated

Secure, fully-isolated AI infrastructure

For a security-sensitive organization: architected and specified a zero-egress AI enclave, with private networking end to end, an MLOps foundation, self-hosted LLMs, and vision object detection. Built and demonstrated a fully isolated, self-contained AI assistant: chat with organizational knowledge, retrieval over internal documents, deep-research workflows, document and diagram generation. Everything runs inside the closed network. Nothing leaves it.

  • Air-gapped
  • LLM
  • RAG
  • Vision
  • MLOps

Built and validated on real data

Multi-region demand forecasting for a global water-technology manufacturer

Built and validated quarterly product-family demand forecasts across five sales regions (EU, UK, US, Australia, China), unifying direct sales with inter-company stock transfers, extended into customer churn analysis and geographic sales analytics. Validated on real multi-year order data.

  • Time series
  • Demand planning

Built and validated on real data

Commodity-price forecasting engine for feed procurement

Built a GPU-accelerated forecasting engine for agricultural commodity prices (wheat, corn, oils): deep time-series architectures (NBEATS, TFT, TiDE and others) ensembled with gradient-boosted models, SHAP-based explainability, and domain feature engineering grounded in feed-formulation chemistry. Validated against real feed-mill purchasing history, aimed at optimal buy-timing and stockpiling decisions.

  • Time series
  • Deep learning
  • Explainability
  • Decision support

Designed and architected

Item-level inventory intelligence for ERP environments

Designed and specified item-level demand forecasting with dynamic safety stock and purchasing recommendations, architected as a cloud API integrating into Priority ERP, for importers and distributors in the $40M to $115M range. Success KPI defined up front: at least a 10% reduction in purchasing error.

  • Time series
  • Systems design
  • ERP

Designed

Three-tier procurement forecasting for construction supply

Designed a progressive-accuracy system for a US construction-procurement SaaS: sales forecast, then supplier lead-time forecast, then dynamic reservation of safety stock, with ERP-integrated automation.

  • Time series
  • Supply chain
  • SaaS

Designed

Receivables anomaly detection for perishable-goods distribution

Designed for a produce wholesaler serving 1,000+ customers: automatic detection of overdue-payment anomalies with weekly alerting, plus daily monitoring of customers who did not order today, integrated with the company's ERP and ordering systems.

  • Anomaly detection
  • Operations

Designed and architected

Freight and currency forecasting for import decisions

Designed and specified a forecasting tool for sea-freight prices and exchange rates as procurement decision support for a plastics manufacturer.

  • Time series
  • Finance

Built and validated on real data

Cashflow prediction from raw banking data

Built and validated a business-cashflow forecasting model working directly from real bank-account transaction history (debits, credits, checks), on an ensemble of deep and classical time-series models.

  • Time series
  • Finance

Designed

Document intelligence for investment decisions

Designed an ML pipeline extracting structured features from scanned business documents to score venture viability (a go or no-go decision plus expected-yield classification) for a real-estate investment group.

  • Document AI
  • Classification

Designed and architected

Generative AI on a historic scholarly corpus

Designed a generative chatbot grounded in the complete writings of a major historic scholar, including the expert-evaluation methodology for domain-authority review and a roadmap to a natural-speech museum installation. The approach was explored with a working local build.

  • LLM
  • RAG
  • Cultural heritage

Built

Behavioral prediction at iGaming scale

Built engaged-user and high-value-player classifiers over hundreds of thousands of betting events, with disciplined evaluation: train, validation, and test splits with leakage control, and full precision and recall reporting.

  • Classification
  • Large-scale tabular ML

Built

Production platform engineering with AI at the core

Built a full production web platform (cloud-native, Kubernetes, a multi-panel product with search, quoting, CRM, and analytics) using an agentic AI development methodology: architecture decision records, automated verification, and AI-driven engineering workflows end to end.

  • Production systems
  • Full-stack
  • AI engineering

Built

Voice-AI agent that converts website visitors

Built a voice AI agent that ingests a website's content through its sitemap, builds product knowledge automatically, and holds spoken conversations that qualify and convert visitors into leads.

  • LLM agents
  • Voice
  • Web

Built and demonstrated

Automated segmentation and measurement for aerial imagery

Built and demonstrated a SAM-based system that automatically segments airfields in satellite and aerial imagery and measures them with precision tooling, wrapped in a purpose-built UI with manual correction, object addition, and report generation. It turns the costliest phase of geospatial vision work into a supervised, minutes-long task.

  • Vision
  • Geospatial
  • Segmentation
  • ML tooling

Built and demonstrated

Organizational knowledge agent for a multi-site operations company

Built and demonstrated a conversational knowledge agent giving a multi-site operations company's staff free-language access to organizational knowledge, demonstrated on real content.

  • LLM
  • Knowledge systems
  • Operations

Built

Conversational operations agent in WhatsApp

Built a full operations agent living in WhatsApp, working in Hebrew voice notes and text. It knows the organization's content library, finds the right videos, manuals, and deliverables, handles compression and format constraints automatically, and delivers them in chat instantly. It clarifies on demand and supports selling, marketing, and in-meeting moments in real time.

  • LLM agents
  • Voice
  • Operations

Advisory engagements

We have also consulted and advised on:

  • AI workflow characterization for a services firm (Tel Aviv and New York)
  • A data and AI opportunity roadmap for a consumer-products company: forecasting, supplier scoring, field inventory, and chatbot analytics
  • AI product strategy for a food-retail concept
  • Agricultural planning, procurement optimization, and a vision-based counting roadmap for a plant nursery