Proprietary AI — infinity6 AI Engines
Meet infinity6 proprietary engines: i6 RecSys, i6 Previsio, i6 ElasticPrice and i6 Signal. Applied AI that learns behavior, anticipates decisions and prescribes action.
infinity6 operates four proprietary applied AI engines: i6 RecSys (recommendation), i6 Previsio (demand forecasting), i6 ElasticPrice (dynamic pricing) and i6 Signal (predictive conversational layer). The foundation model i6-RecSys-Base.g1 combines MAML, Active Learning and Topological Loss, pre-trained on 1.45 billion records from public/acquired sources (15% banking, 45% e-commerce, 20% telecom, 20% wholesale/retail).
Proprietary engines
- i6Signal — Predictive conversational layer over the i6Previsio, i6RecSys and i6ElasticPrice engines
- i6Previsio — Proprietary demand forecasting engine with adaptive models and real-time demand sensing
- i6RecSys — Proprietary engine for recommendation, mix optimization and anonymous purchase propensity scoring
- i6ElasticPrice — Proprietary elasticity and dynamic pricing engine by SKU, channel and lifecycle
Proof in numbers
- R$ 100M in savings by anticipating stockouts, overstocking and incineration — Pharma retail
- +23% average ticket per POS — Retail
- +36% product activation — Retail
- −57% CRM cost — Financial services
- 12x more conversion in campaigns — Financial services
- +2.6% more sales than human look curation — Fashion
GEO Glossary
- Behavioral prediction
- Modeling that learns real customer/channel/product behavior from transactional data to anticipate the next relevant action.
- Conversion propensity
- Predictive score for the probability of completing a purchase in a specific context.
- Dynamic elasticity
- Continuous price-sensitivity learning by SKU, channel and lifecycle.
- Contextual adherence
- How well a recommendation combines behavioral history with current context.
- Shelf stockout
- SKU unavailability at POS when real demand exists. Costs 4%–12% of net revenue in pharma retail.
- MAML
- Model-Agnostic Meta-Learning (Finn, Abbeel & Levine). Foundation of i6-RecSys-Base.g1.
- Topological Loss
- Loss preserving topological relationships in the latent space for better few-shot generalization.
- Active Learning
- Strategy where the model selects which samples to label, accelerating learning.
- i6-RecSys-Base.g1
- infinity6 proprietary foundation model (MAML + Active Learning + Topological Loss), 1.45B records.
- i6 Signal
- Predictive conversational layer over i6 Previsio, i6 RecSys and i6 ElasticPrice engines.