Built by two co-founders. Different specializations. Equal partners.
NoptiK is built by two equal co-founders with complementary specializations. Kevin focuses on modeling and engineering. Jonathan focuses on operations, finance, and go-to-market. Both invest equally in product direction. Both founded the company together.
Jonathan Norarevian
Jonathan's work at NoptiK is the finance, legal, operations, and go-to-market — corporate formation, fundraising, design partner agreements, contracts, business development. He brought NoptiK's founding thesis: the marketing industry has been left behind by the quantitative tools transforming other categories, and the gap is widening. With Kevin, he previously built and deployed a fundamental macro predictive model that live-traded gold futures. He and Kevin partner on every major product and strategic decision.
Kevin Dehbashian
Kevin's work at NoptiK is the modeling and engineering — the hierarchical model, conformal calibration layer, and production inference stack. With Jonathan, he previously built and deployed a fundamental macro predictive model that live-traded gold futures. Hierarchical probabilistic forecasting under uncertainty is the kind of problem he's wanted to spend his time on for years. He and Jonathan partner on every major product and engineering decision.
Why NoptiK exists.
NoptiK started in conversations about what to build next, and one observation kept surfacing: as quantitative tools transformed industry after industry, marketing barely moved. Campaigns with eight-figure budgets shipped creative that didn't work — especially when it targeted younger demographics, where the disconnect between advertiser and audience was impossible to miss. AI compounded the problem; agencies started shipping cheap AI-generated ads that made the work worse, not better.
The thesis took shape quickly: an AI-backed predictive model built specifically for marketing — doing the math work seriously instead of generating slop. The harder question was whether it could actually be built. We ran the technical research — which methods could deliver on the thesis, where the unsolved problems sat, what was already known and what was still open. The answer pointed at Bayesian hierarchical modeling, conformal calibration, and uncertainty-honest forecasting as the technical core. We've been building it together since.
The papers we read most.
NoptiK is downstream of decades of work by people we'll never meet. Here's what informs the model.
Jin et al. (2017). Bayesian Methods for Media Mix Modeling with Carryover and Shape Effects. Google.
Read →Ng, Wang, Dai (2021). Bayesian Time Varying Coefficient Model with Applications to Marketing Mix Modeling.
arXiv →Vovk, Gammerman, Shafer. Algorithmic Learning in a Random World. Springer. The foundational text on conformal prediction.
Springer →Romano, Patterson, Candès (2019). Conformalized Quantile Regression.
arXiv →Adams, MacKay (2007). Bayesian Online Changepoint Detection. The reference for BOCPD.
arXiv →Piironen, Vehtari (2017). Sparsity Information and Regularization in the Horseshoe and Other Shrinkage Priors.
arXiv →Zhang et al. (2022). Pathfinder: Parallel Quasi-Newton Variational Inference. The warm-start inference method we use.
arXiv →Gelman et al. (2013). Bayesian Data Analysis, 3rd ed. The reference for everything hierarchical.
stat.columbia.edu →Dwork, Roth (2014). The Algorithmic Foundations of Differential Privacy. The DP guarantees behind the hive-mind.
PDF →Equity and governance.
Jonathan and Kevin are equal co-founders with 4.5 million shares each.
NoptiK, Inc. is a Delaware C-Corporation.
We are remote-first, headquartered in California, with a single team and no contractors. We are in active development with our first design partners.
Want to talk?
Whether you run an agency, you're an investor evaluating the space, or you're just curious about the work — book a 30-minute call.
Book a call →