Machine Learning Research Engineer

About Parisi Labs

Parisi Labs is an AI company building learning systems for complex physical environments. We combine historical and live data with real operational context to help people understand the present, evaluate possible futures, and make better decisions.

Energy is our first proving ground. Ask The Grid (https://askthegrid.com) is our public product for exploring the systems, markets, and assets that make up the power grid. We are a small technical team working across machine learning, data infrastructure, software, and real-world operations.

About The Role

We are looking for a machine learning research engineer to work directly with our Chief Scientist and accelerate our core modeling work.

You will inherit a real model and evaluation system, understand how it behaves, and make it materially better. That means implementing ideas from papers, designing careful experiments, debugging training and data problems, improving evaluation, and translating successful research into reliable systems.

This is neither a purely academic research position nor a conventional production-ML role. It is for someone who enjoys the full empirical loop: form a hypothesis, build the experiment, determine whether the result is real, and ship what works.

What You Will Own

- Reproduce, extend, and improve our model-training and evaluation systems.

- Design experiments and ablations that separate meaningful improvements from noise, data problems, and evaluation artifacts.

- Investigate model behavior through error analysis, diagnostics, and carefully constructed benchmarks.

- Build better tooling for experimentation, tracking, reproducibility, and technical decision-making.

- Work closely with data and product engineers to turn research requirements into dependable systems.

- Translate promising research into production-quality implementations.

- Communicate results clearly: what changed, what the evidence shows, and what we should try next.

- Help establish the research practices and technical standards of an early AI company.

First 90 Days

- 30 days: Reproduce the current model and evaluation system, identify fragile assumptions, and ship an early improvement to the research workflow.

- 60 days: Own an experiment from hypothesis through implementation, evaluation, and failure analysis.

- 90 days: Run a dependable weekly research cadence with reproducible results, clear readouts, and evidence-backed recommendations.

You May Be A Fit If

- You have an MS, PhD, or equivalent demonstrated depth in machine learning, computer science, statistics, applied mathematics, electrical engineering, or a related field.

- You can read a paper, implement the important idea, and determine whether it actually works.

- You have strong Python and modern machine-learning framework experience.

- You understand experimental design, statistical reasoning, and the many ways an ML result can be misleading.

- You have worked with sequence models, probabilistic modeling, forecasting, scientific ML, optimization, or other learning problems grounded in real systems.

- You have improved a real model under practical data, compute, or deployment constraints.

- You write clear research code and communicate technical conclusions without hiding behind jargon.

- You want substantial ownership and can operate without a large, mature research organization around you.

A particularly strong archetype is someone with a research-heavy graduate background followed by two or three years of applied industry work, but credentials are not a substitute for evidence of excellent work.

Helpful Background

Location And Working Style

Boston/Cambridge is strongly preferred. New York City can work for an exceptional candidate with a regular in-person cadence.

Compensation And Benefits

Base salary range: $210K-$275K, plus meaningful early-stage equity, medical, and dental benefits. Final compensation depends on level, location, experience, and role scope.

Interview Process

- Conversation with the Chief Scientist.

- Research working session or compact experiment and evaluation review.

- Technical calibration with the CTO.

- In-person final in Boston/Cambridge or New York City.

- Offer review.