# Hiring: Founding Bayesian / ML Engineer — MMM platform (remote EU, equity + cash)

**URL:** <https://discourse.pymc.io/t/hiring-founding-bayesian-ml-engineer-mmm-platform-remote-eu-equity-cash/17700>\
**Category:** Jobs\
**Tags:** modeling, pymc-marketing\
**Created:** [April 25, 2026, 3:35pm UTC](https://discourse.pymc.io/t/hiring-founding-bayesian-ml-engineer-mmm-platform-remote-eu-equity-cash/17700 "2026-04-25T15:35:26Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![CalybraFounder](https://avatars.discourse-cdn.com/v4/letter/c/977dab/32.png) [@CalybraFounder](https://discourse.pymc.io/u/CalybraFounder)\
**Post date:** [April 25, 2026, 3:35pm UTC](https://discourse.pymc.io/t/hiring-founding-bayesian-ml-engineer-mmm-platform-remote-eu-equity-cash/17700/1 "2026-04-25T15:35:26Z")

</div>

Pre-seed startup in Europe building a hierarchical Bayesian Marketing Mix Modeling platform. Looking for a founding engineer who has shipped serious probabilistic programming to production, ideally with hierarchical models on commercial / marketing data. Remote EU (worldwide with EU overlap negotiable). **Co-founder-level equity + cash commensurate with experience.** Apply: [hello@calybra.to](mailto:hello@calybra.to) — link to a PyMC/Stan project you’re proud of, plus a 3-min Loom on a Bayesian decision you made in production.

* * *

**What we’re building.** A platform that does MMM for multi-channel brands spending €3M+/year across paid media, TV, OOH, and offline. MMM is the right tool for the job — platform-reported attribution has been broken for years and is now getting worse with cookie deprecation, DMA, and AI-managed buying (Performance Max, Advantage+) — but most MMM in-market today is still delivered as consultant-led quarterly PDFs. We think the category deserves a modern product with a proprietary modeling engine, validated calibration, response curves and scenario planning surfaced directly to the marketing buyer, and analyst-in-the-loop review (we treat “a human approves the model” as a feature, not a limitation).

**Prior art we’ve read and respect.** Meta Robyn, Google Lightweight MMM, Google Meridian, Jin et al. (2017) on Bayesian methods for media mix modeling, and the broader literature on geo-based causal inference and hierarchical shrinkage in marketing. We’re not reinventing adstock and saturation — we’re trying to productize the best of what’s out there, make it auditable for a non-technical buyer, and close the loop with incrementality tests where clients can run them.

**What you’d own.**

- The core hierarchical Bayesian model: priors, group structure across clients/categories, adstock and saturation parameterization, sampler choice (we’re PyMC-first, open to numpyro or Stan where it helps).

- The validation pipeline: geo-holdout, posterior predictive checks, calibration against lift tests, identifiability diagnostics.

- The data schema and ingestion contract with our product engineer (Meta Ads, Google Ads, GA4, TikTok connectors + CSV).

- The “model confidence” surface the customer sees in the dashboard — this is our trust-building promise, not a nice-to-have.

**Who we’re looking for.**

- You’ve built Bayesian models in production using PyMC, Stan, or numpyro. You can talk concretely about a time you changed a prior and it actually mattered.

- You have opinions about identifiability and you’ve debugged it in a real hierarchical model with messy data.

- You’ve validated models where there is no ground truth. You know the difference between “fits the data” and “decision-relevant”.

- Bonus: MMM, media attribution, retail/CPG forecasting, or hierarchical econometrics experience. Contributions to PyMC, Robyn, LightweightMMM, or Meridian are a strong signal — we’d love to hear about them.

- Nice-to-have: comfort shipping code (not just notebooks), interest in working closely with a product engineer, and curiosity about the commercial side (why the CMO will use this).

**Not what we need.** A generic ML engineer who did deep learning and is curious about Bayesian. A researcher who doesn’t want to see their work in a customer’s hands. Someone who thinks an LLM can replace the model (it can’t — but it can make the output easier to act on).

**Compensation.** Co-founder-level equity + cash commensurate with experience. Standard 4-year vesting with 1-year cliff. We’re a small team, remote EU, full-time. Happy to discuss the cash/equity trade-off openly on the first call.

**How to apply.** Email **[hello@calybra.to](mailto:hello@calybra.to)** (or DM me here on Discourse) with:

1. Link to a repo, notebook, or OSS contribution you’re proud of.

2. 3-minute Loom (or audio file) on the most important Bayesian decision you’ve made in production and why.

3. CV optional.

We’ll reply within 5 working days, whatever the outcome.

Happy to answer technical or process questions publicly in this thread — assume anything asked here is useful to other candidates reading it later.

-– _The Calybra team_

_A note on who’s behind this. Calybra is led by a serial founder and investor with a track record of building and backing successful companies. We mention it because at pre-seed there isn’t much else public about us yet, and we know that matters when you’re considering a founding role._

---

<div class="post-metadata">

**Author:** ![anevolbap](https://avatars.discourse-cdn.com/v4/letter/a/9de0a6/32.png) [@anevolbap](https://discourse.pymc.io/u/anevolbap)\
**Post date:** [June 16, 2026, 1:49pm UTC](https://discourse.pymc.io/t/hiring-founding-bayesian-ml-engineer-mmm-platform-remote-eu-equity-cash/17700/2 "2026-06-16T13:49:44Z")

</div>

@CalybraFounder, looks very interesting. I wanted to check whether the position is still open.