Turn Business Data into Measurable Value with Practical AI/ML Systems
I help growing companies turn existing data into better forecasts, smarter decisions, and working AI products — from identifying the opportunity and building the models to deploying systems that deliver results.
What I Help With
Forecasting & Predictive Modeling
Anticipate demand, operational risk, and rare events with models evaluated against real-world outcomes.
Decision Science & Optimization
Turn forecasts and business constraints into better pricing, inventory, prioritization, and resource-allocation decisions.
AI Search, RAG & Hybrid Retrieval
Make complex document collections searchable using semantic retrieval, BM25 keyword search, reranking, and grounded AI workflows.
Production ML & AI Systems
Build reliable data, modeling, monitoring, and retraining workflows that work beyond the prototype stage.
Why This Matters
A useful AI/ML system does more than produce a prediction or a demo. It solves a specific business problem, is evaluated against meaningful outcomes, respects operational constraints, and can be maintained as data and needs change. That’s the connection I focus on: from data and models to decisions and working systems.
Featured Work
Diagram coming
Pricing & Optimization Systems
At Thrasio, I developed production pricing and inventory decision systems combining demand forecasting, price elasticity, and financial constraints. An inventory-aware pricing initiative increased profitability by 15% while reducing excess inventory by 20%. Separate elasticity work covered 10,000+ products.
Diagram coming
Predicting Housing Complaints Before They Occur
Built a quarterly risk model using NYC housing data across buildings, apartments, and complaint categories. A retrospective evaluation of roughly 7 million previously generated scores found a 6.3× complaint-rate lift in the highest-risk 10% — evidence of useful risk ranking, not a claim of measured operational savings.
Diagram coming
RAG & Hybrid Search Product
Developed and evaluated document retrieval for a collection of approximately 10,000 documents, benchmarking three embedding models. The hybrid search design combines BM25 keyword retrieval, semantic vector search, and reranking to surface more relevant source material.
Ways to Work Together
Project-Based AI/ML Consulting
Build or improve a defined forecasting, optimization, retrieval, or production ML system — with clear deliverables and evaluation criteria.
Technical Audit & Advisory
Get an independent assessment of data readiness, model quality, RAG performance, technical risks, or the next step for an existing prototype.
Fractional AI/ML Leadership
Get senior technical direction on strategy, architecture, validation, and delivery without a full-time AI/ML leadership hire.
About
I’m Isroel Kogan, an AI/ML engineer and consultant with experience spanning pricing and inventory optimization at Thrasio and production ML and AI search at Cadmus. Through Kogan AI Consulting, I bring business-focused modeling and hands-on engineering together to help teams build useful systems from their data.
Have a Similar Problem?
If you have useful data, an uncertain model, or a workflow that could benefit from smarter decisions, let’s discuss what’s practical and worth building.