SmartInferCase Studies
SmartInfer is built from two decades of production work in search, ranking, recommendations, experimentation, document AI, and enterprise engagement systems. These case studies show the operating patterns behind the company: diagnose the real failure, separate data and system problems from model problems, and build ML systems that move business outcomes.
Why these matter
-
From search failures to business intelligenceThe same patterns recur across retail and enterprise AI: customer intent is hidden in queries, conversations, content gaps, failed recommendations, and experiments. SmartInfer turns those signals into product discovery, business memory, and decision intelligence.
Commerce search, ranking & discovery
-
60% of search failures were business problems misclassified as technology problems. A diagnostic framework that separated relevance, ranking, and demand–supply gap issues before building models.
-
Replacing a 15-year heuristic ranking system with ML. The key was extracting tacit knowledge from the heuristic into features and designing price-normalized training targets.
-
Price defines consideration sets, conditions relevance, and determines marketplace value allocation. Three interventions — each producing >2% revenue lift — that form a structural hierarchy every commerce ML system should address.
-
Boosting high-quality images in search made pages look better — and revenue dropped. In core C2C categories, low-quality mobile photos are authenticity signals. Buyers optimize for trust, not aesthetics.
-
Using topic models to identify query spaces where customer demand exists but catalog supply does not — closing the gap between what customers search for and what the marketplace offers.
Experimentation & analytics
-
A/B testing in e-commerce search: visit-, query-, and item-level bias; OEC selection; non-parametric testing for skewed metrics; holiday seasonality; and budget constraints.
-
The measurement and analytics framework behind the diagnostic approach — how to instrument a search system so failures are classified before resources are committed.
-
A/B testing in a marketplace where treatment on one side affects outcomes on the other. Designing experiments that account for interference, network effects, and equilibrium shifts.
Enterprise AI & document understanding
-
Replaced GPU-bound deep learning with distributed Spark, built Bayesian ranking so marketers could search by campaign intent, and worked with pricing to hit 75% margin — scaling to 5B profiles while cutting cost by over 70%.
-
A document understanding system stuck at 20% precision wasn’t a model problem — it was a data problem. First-principles diagnosis, systematic data-centric interventions, and iterative error analysis tripled precision.
-
Diagnosed four failure modes and built four coordinated systems: feed relevance ranking, multi-entity recommendations, Q&A answer ranking, and spam detection.
-
Customers couldn’t configure chatbots; agents didn’t trust answer suggestions. Two different failures requiring two different fixes — vertical defaults for the chatbot, dual-index retrieval for the answer engine.
-
Mathematical formalization of the unified latent space, composite engagement score, and two-stage adaptive target optimization for multi-entity recommendations across heterogeneous organizations.
Additional case studies on AI memory, agent verification, marketing science, and retail intelligence are in preparation. The full archive lives at anjangoswami.com/case_studies.