Quantitative Developer – Statistical Modeling & Analytics

Location: Remote (Latin America) | Type: Full-Time

We're a performance-driven digital marketing agency building proprietary software at the intersection of Google Ads, data analytics, and AI. Our platform turns raw advertising data into decision-grade signals across a portfolio of managed accounts. The product is live and actively used in production.

This is early-stage product work, and we're expanding the team deliberately: hiring for complementary skill sets, not redundancy.

We already have a strong engineer on the team with deep LLM, RAG, and systems-building experience. This hire is intentional and complementary: we're looking for someone whose center of gravity is quantitative, not software-first — a person who thinks in estimators, distributions, and model assumptions, and who can also ship that thinking as production code.

You'll work directly with the founder and the existing engineer to own the analytical engine of the platform: the statistical logic that decomposes performance, detects real change from noise, and explains why metrics moved. Where the existing engineer builds the AI and evaluation layer, you build the quantitative foundation underneath it.

This is not a pure data-science role — you'll write real product code in TypeScript and ship features in Next.js on a Supabase backend. But the reason we're hiring you is the math.

Key Responsibilities

• Design and implement the analytical methods behind the platform: log-space decomposition of performance metrics, rolling/windowed regression for elasticity and trend estimation, and change-point detection (e.g. CUSUM, PELT) to separate genuine shifts from noise.
• Build anomaly detection that respects the realities of advertising data — non-stationarity, autocorrelation, fat tails — rather than assuming clean, normally distributed inputs.
• Develop the quantitative logic for our "explain what changed" feature: quasi-experimental and counterfactual reasoning on observational data, where there is no clean control group and no ground truth. This is the hardest and most important part of the role.
• Reconstruct joint distributions from marginal data (e.g. iterative proportional fitting / raking) for demographic and segment-level analysis.
• Build benchmarks, baselines, and forecasts across key advertising metrics, and translate them into clear internal signals.

• Define, statistically, what "good" and "reliable" mean for AI-generated insights — the rigor that the existing engineer's evaluation framework measures against. (You provide the statistical definitions and validation logic; you are not the owner of the LLM eval framework itself.)
• Integrate the Claude (Anthropic) API to turn validated quantitative findings into insight summaries, performance narratives, and recommendations.
• Contribute to prompt design and output checks where they intersect with the statistical layer — ensuring narratives are grounded in the numbers and don't overstate significance.

• Build and maintain features across the stack using TypeScript, Next.js, and Supabase (PostgreSQL, auth, storage), deploying via Vercel.
• Integrate with the Google Ads API, writing and optimizing GAQL queries against large, messy production data.
• Implement and maintain OAuth flows within the Google ecosystem.
• Collaborate with the existing engineer on architecture and technical tradeoffs.

• Design data models and query logic optimized for analytical workloads on PostgreSQL.
• Build internal dashboards and reporting tools that surface statistical insight clearly and honestly (variance, not just averages).
• Keep data pipelines reliable, reproducible, and documented.

• A genuinely strong quantitative foundation — demonstrated through a background in statistics, econometrics, applied math, physics, operations research, or quantitative finance, or equivalent applied modeling work. We weight depth over years.
• Fluency in the methods that actually matter here: regression and time-series modeling (including awareness of multicollinearity, heteroskedasticity, and stationarity), change-point and anomaly detection, and causal / quasi-experimental inference on observational data.
• Experience modeling on large, noisy, real-world datasets — not just clean academic or Kaggle data.
• Production proficiency in TypeScript — or a proven ability to prototype in Python/R and port cleanly to TypeScript without supervision.
• Strong SQL and PostgreSQL (Supabase or comparable backend-as-a-service experience preferred).
• Experience building or consuming REST APIs and deploying via Vercel or similar.
• Experience integrating an LLM API into a production workflow
• Comfortable working independently in a fast-moving, founder-led environment with no fixed playbook.

• Direct experience with the Google Ads API / GAQL or other advertising-platform APIs.
• Background in marketing analytics, ad tech, or performance-marketing data.
• Experience with statistical computing libraries (NumPy/SciPy/statsmodels, R, or JS-ecosystem equivalents).
• Experience building internal analytics tools or SaaS products.
• Exposure to other Google integrations (Analytics, Merchant Center) 
• Startup or early-stage product experience.

You are a quantitative thinker who ships production code. You write clean, maintainable TypeScript and can deliver features without hand-holding, but what sets you apart is how you reason about data.

Before drawing a conclusion, you ask whether the difference is statistically meaningful. You model variance, not just the mean. You've worked with messy, real-world data and you know the difference between a signal and an artifact. When there's no ground truth, you reach for a defensible inferential approach rather than a correlation dashboard.

You complement a strong engineering partner rather than compete with one. Where they build the AI and systems layer, you build the quantitative understanding it stands on.