Probabilistic demand forecasting. GenAI product content. Agentic merchandising. Semantic and agentic search. Personalization. Store-ops computer vision. Built in your stack, against your KPIs.
Each is a system we've shipped or have direct shipped credibility for. We don't take engagements outside this surface area unless we can name an adjacent build we've already done.
SKU-store-week probabilistic forecasts that drive allocation, replenishment, and open-to-buy. Recommendation systems shipped across thousands of stores with department-level precision.
PRIMARY KPI · On-hand · Sell-through · OOS rate
Product copy, attribute extraction, multilingual content, and merchandising-grade GenAI with brand-voice grounding. Eval pipelines that catch hallucinations before they hit production.
PRIMARY KPI · Content velocity · Quality acceptance
Agent workflows for assortment review, markdown recommendation, and exception management. Built with deterministic policy layers and merchant-in-the-loop escalation.
PRIMARY KPI · Margin · Markdown depth
Semantic search, query understanding, and conversational discovery for e-commerce — built for retrieval quality and conversion, not just relevance scores.
PRIMARY KPI · Conversion · Search abandonment
sales · inventory · calendar
online / offline
Spark-scale
Databricksfine-tuned
action
Representative builds — the technology patterns we bring to retail. Where we cite results, they’re from work our team has shipped; everything else is what we’d build for you.
RT.T1 · Databricks
Spark-scale hierarchical forecasting fine-tuned per store and department — the muscle behind a 3,000+ store recommendation system at ~82% precision.
Stack · Databricks · MLflow · feature store
RT.T2 · LLM
Attribute extraction and copy generation integrated with PIM, brand-voice-constrained with eval gates.
Stack · Claude/GPT · eval harness · PIM integration
RT.T3 · Agents
Agentic workflows over governed retail data — markdown recommendations, assortment queries in natural language, exception surfacing.
Stack · LangGraph · Databricks · governed retail data
Retail is less regulated than insurance or healthcare, but operationally unforgiving. Our governance frame is the P&L — gross margin, on-hand, sell-through, markdown depth — not a regulator's checklist.
Where regulation does apply (consumer privacy, retail media measurement, generative ad content), we apply the same audit-readiness discipline we use in insurance and healthcare.
Our founding team shipped AI inside a Fortune-100 retailer's merchandising and supply-chain organization, including a SKU-store-week recommendation system fine-tuned per store and department across thousands of locations.
Most first conversations are 30 minutes with a senior partner. We'll tell you whether it's a fit, whether AI is the right answer, and what we'd build first.