New York, NY Summer 2026 10–12 weeks $60–80 / hr

Research Intern, Althea Research

Work directly with the President on a frontier vulnerability-modeling problem with real underwriting consequences.

Most homes in America are insured against catastrophe by carriers that cannot tell the difference between a house that will survive the next hurricane and one that won't. The vulnerability models are too coarse, the data is too aggregated, and the feedback loop between what actually happens in a storm and how the next policy gets priced is broken. The result is a market that retreats from risk it cannot measure, instead of pricing the risk well enough to reward the homeowners who build for resilience.

Althea Research is the frontier research arm of Althea. We build the models that close that gap. Vulnerability at the individual asset. Resolution at the level of a specific roof in a specific storm. Pricing accurate enough to make resilience the economically rational choice for the homeowner, and underwriting capacity that grows as the data spine compounds. The work is intellectually serious and operationally consequential. Every model we ship changes what gets built, what gets insured, and what stays standing.

We're hiring one research intern for summer 2026 to work directly with the President on a problem at the frontier of that program.

The problem

Climate is interacting with the physical world in new ways, and homes are paying the price. Storms are stronger. Hail is bigger. Wildfires reach further. The homes built in the last fifty years were built for a different climate, and the insurance system built around them was built for a different loss profile.

The work is to understand those risks at the level of the individual home, and to help homeowners reduce them. Which roofs hold under which winds. Which trees fall on which structures. Which mitigation choices actually change the loss outcome and which only look like they do.

Get the physics and the data right, and the rest follows: better pricing, better incentives, more homes that stand up to the next storm.

That is what Althea Research exists to do.

The team you're joining

Althea Research is small, deliberately. The model is closer to a quantitative trading research desk than a corporate R&D function. Eventually two to three senior researcher-engineers reporting to the founders, supported by interns. This intern is the first research hire. The function is being built around the work, not the other way around.

Six to twelve months ahead of what the operating business currently needs. Close that gap and the team is behind. Push past it and the team has drifted out of relevance.

The team's posture is forward-deployed. Althea Research runs six to twelve months ahead of what the operating business currently needs, building the models the MGU's underwriting will rely on next, identifying the next axis of resolution along which the market is blind, replacing external dependencies with proprietary capability before the dependency becomes painful.

The intern reports to the President and works closely with the engineering team building the core product. This is not a structured program. There is no rotational track, no curated intern project, no shadowing component. It is a 10–12 week appointment for someone who wants to ship something testable against a real underwriting decision, and who wants their work to help shape what the research function becomes.

What you'd work on

The Althea Research agenda spans four areas. The summer's specific problem will be defined in conversation, based on your background, our priorities at the start of the summer, and what is tractable in 10–12 weeks. Below are the live directions and the texture of the work in each.

Refining the construction-class vulnerability curves that are standard in cat modeling into something that resolves at the individual asset. Combining structural engineering priors with empirical loss data and engineering test data. Operationalizing IBHS field research into curves that an underwriter can use to price a specific home.

Physics-informed statistical models FORTIFIED™ designation integration Roof-system decomposition Hail vulnerability Boundary-layer wind loading

Extracting underwriting-relevant features from imagery at population scale. Roof geometry, roof condition, secondary features (solar, trampolines, pools), tree overhang, drainage. The methodological problems are real: label scarcity, geographic distribution shift between training and inference, calibration when ML-extracted features feed into regulated rate filings.

Aerial imagery Computer vision Active learning Calibration Distribution shift Pre/post-event imagery

Refining hazard footprints at higher spatial resolution than vendor models provide. Damage attribution that combines pre-event imagery, post-event imagery, and claims data to produce calibrated probabilities rather than binary determinations. Tree fall and wind-borne debris models that operate at portfolio scale.

Tropical cyclone reconstruction Hail swath modeling Tree fall models Wind-borne debris Post-event imagery analysis Attribution models

Translating asset-level vulnerability into portfolio-level capital requirements. The dynamic capital loading that prices the marginal contribution of an individual asset to portfolio tail risk. Correlation structure as a function of construction standard mix. Adverse selection inversion under third-party certification.

Event-level loss tables Dynamic capital loading Tail correlation Marginal PML contribution Adverse selection modeling

What we're looking for

We care about how you think and what you've built, not the credentials on top.

Master's or PhD candidate in a quantitative field

Applied math, statistics, computer science, machine learning, physics, operations research, economics, engineering. Year in program doesn't matter. We do not require, and do not particularly prefer, formal training in wind engineering, structural engineering, or insurance. Domain expertise is acquirable. The thinking is not.

First-principles quantitative orientation

The lineage we recruit from is closer to Jane Street, Hudson River Trading, D.E. Shaw, DeepMind, and Anthropic Research than to insurance or cat modeling. If you have spent time in or near that pool, through internships, research, or work that engaged with that style of thinking, that is the signal we are reading for. If you haven't, your CV should make the case some other way.

Shipped output, not academic theater

The work product at Althea Research is testable models deployed against specific underwriting decisions, benchmarked against the legacy approach they replace. Not papers. Not frameworks. Not whitepapers. You should be comfortable with the unit of work being something that runs, makes a prediction, and can be measured against a baseline by the end of the summer.

Business curiosity

You read the research direction and wonder why we framed it that way. You ask how the model will be used and let that shape the methodological choices. You notice when a technical decision has a business or regulatory implication that changes the right answer. You write your own spec when the one we gave you turns out to be wrong.

AI fluency

You use modern ML tooling and agent-assisted coding as a core part of how you work. You have a point of view on where these tools belong in research workflows and where they introduce risk. You stay near the current edge.

Fast learner

Insurance is a new domain for most people we recruit. We don't expect you to arrive knowing it. We do expect you to learn it quickly, take it seriously, and let it shape how you build.

Why it matters

The work at Althea Research is the most leveraged technical work in the company. The MGU writes policies. The research arm decides what the MGU is capable of writing in eighteen months. Every model that ships changes the addressable book, the loss ratio, the capital efficiency, and the terms we can negotiate from reinsurance and capital partners. Ten weeks of focused work on the right problem can move the institution's trajectory.

It also matters outside the company. The reason climate volatility is producing an insurance retreat is that the industry's modeling infrastructure was built for a different era and cannot be retrofitted in time. Building the replacement is one of the highest-leverage interventions available against the structural problem in property insurance. If Althea is successful, the result isn't just a strong company. It's a structural shift in what gets built and retrofitted, a world where more homes remain insurable because they actually hold up.

That's the opportunity: a hard technical problem, a real operational consequence, and a lever that changes the world.

The Althea Person

This interdisciplinary work depends on having the right people. We are actively assembling a team, a board, and a cap table. Maybe you belong at Althea.

About Althea

With Althea, insurance becomes an investment in resilience rather than a tax on bad luck. We make insurance more affordable for homeowners who build stronger roofs. We exist to help our homeowners weather the next storm. And the one after that.

We're well capitalized and tracking to write our first policy in fall 2026.

Geoffroy Bablon
Geoffroy Bablon
Founder & CEO · Capsule founding team
Jordan Breighner
Jordan Breighner
Founder & President · Finance, risk, complex systems
Joe Crobak
Founding Engineer · Confluent, US Digital Service, Foursquare
Marc Ivins
Head of Sales · Semsee, Cover Whale, Swyfft
Roderick Thaler
Roderick Thaler
Founding Chairman & Senior Advisor · Former Vice Chairman, Aon Re

Why physical risk. Why now.

For most of the modern era, catastrophe carriers managed volatility through hazard pricing, exposure management, reinsurance, and rate increases. The math worked. Resilience pricing was a nice-to-have. The tools didn't exist, the carriers didn't need them, and the problem was latent rather than urgent.

Climate volatility has broken that equilibrium. The remaining lever is resilience, and the industry does not have the tools to use it. Hazard and exposure are still well-modeled. Vulnerability is the unsolved problem, and it's the only one that matters now.

Althea Research starts with the roof because that is where the physics are clearest, the data is verifiable, and the leverage on insured loss is highest. The roof is also where the FORTIFIED™ standard has produced a market of verifiable resilient assets that the legacy modelers cannot see.

The expansion path within wind and hail is wide: hurricane, severe convective storms, contractor-level resolution, manufacturer-level resolution, code-enforcement resolution. Each probe is a row in a training dataset no one else can assemble.

1
Verified parcel-level data

Captured through the same network that manufactures the resilient risk: evaluators, inspectors, contractors, claims pathways.

2
Proprietary models

Resolving vulnerability at the individual asset, calibrated against engineering tests, post-event reconnaissance, and accumulating book data.

3
Operational deployment

Through Althea's underwriting, where each policy bound makes the next one more defensible and the homeowner's premium more accurately reflects the resilience of their home.

4
A housing stock that gets stronger

Because the price signal finally reaches the homeowner: better insurance for better-built homes, with the modeling rigor to back the differentiation up.

The work compounds. Each policy bound is a longitudinal sensor on a specific structure: construction spec, exposure, performance under stress, paid loss. The data spine that accumulates from those sensors is what every subsequent model is built on, and what no later entrant can buy.

Compensation

$60 – $80 / hr

10–12 weeks. The range is a guide, not a ceiling. We calibrate to the depth of your craft and the scope of your research agenda.

In-person preferred, remote considered.