Careers
Ten seats. One bench.
Blue Horizon Labs is forming a bench of ten practitioners around the founder: the people the lab would put on a client engagement, plus the operators who keep a lab honest. The shape is set; the seats fill in the order revenue allows, and each one converts from contract or part-time to employee at a gate the person is told before they commit.
How the lab works
Research by building, and the findings published either way.
Research by building
Every engagement is applied research: a diagnostic reads the business first, the system is architected to what it actually needs, and the result is measured against a baseline rather than asserted. The lab runs on the same architecture it installs.
Findings published, including the ones that failed
Studies record a hypothesis, a baseline, a method, and a threshold up front, then publish the outcome against that threshold — confirmed, partial, null, or negative. Internal work is published warts and all; client work is anonymized and non-positive outcomes are published only with consent.
The gate is a test, not a reviewer's memory
Design laws, price claims, client-name rules, and disclosure rules are enforced by scripts that fail the build. If a rule matters, it becomes a test. If nobody will write the test, the rule was decoration.
Small, and honest about it
This is a small operation, and the page says so plainly. The work is real and the systems are in production; nothing here is padded to look bigger than it is.
The ten seats
10 seats, 7 of them on the delivery bench.
- Every seat enrolls in the Anthropic partner certification path as part of onboarding. The lab is open about why: the Claude Partner Network's review counts certified practitioners, and a lab that builds on Claude should have people who are certified on it.
- Seven of the ten are delivery seats, the people a client would meet. Three are the operating bench: growth, operations and finance, content and brand.
- Terms publish here when they are ratified. Until then the seat is a description and a conversation, and the application form is open.
- S1DeliveryForming
Fixed-Ops Delivery Consultant
Runs the fixed-operations review inside a dealership's service department: reads the service drive against a baseline, writes the findings, and leads the thirty-day workflow pilot that follows.
You would own
- The fixed-operations review, from first visit to written readout
- The thirty-day single-workflow pilot and its before-and-after measurement
- The dealer relationship, with the founder, through the pilot
We look for
- Years inside a dealership service department or fixed-operations consulting
- Comfort reading a repair-order report and saying plainly what it shows
- A willingness to publish what did not work, which is the lab's whole research standard
- Engagement
- Contract to start; converts at the first paid dealership review
- Location
- Capital Region, New York; on site at dealerships
- Terms
- Pending ratification
- S2DeliveryForming
Forward-Deployed Solutions Engineer
Sits with the client and builds the agent or the workflow on Claude where the work actually happens, then hands it over with a runbook the client's team can operate without us.
You would own
- The build of each client agent pack, from scope to production on the lab's account
- The runbook and the handover, written so another engineer could take it on
- The weekly cost and usage report for each deployment
We look for
- Shipped software that real people depend on, with the scars to show for it
- Fluency with Claude, tool use, and the plumbing around a worker and a data store
- Patience with a business owner who does not care how it works, only that it does
- Engagement
- Contract to start; converts at the second production deployment
- Location
- Capital Region preferred; remote in the United States considered
- Terms
- Pending ratification
- S3GrowthForming
Growth & Partnerships Lead
Owns the pipeline: the warm-introduction lists, the chambers and the business improvement districts, the partner relationships, and the honest weekly count of conversations that moved.
You would own
- The prospect lists and the outreach the founder sends under their name
- Partner and community relationships across the Capital Region and Hudson Valley
- The weekly pipeline report, measured against the published hypotheses
We look for
- A record of opening doors for a small firm, not a large brand
- Discipline about what counts as a lead and what is a coffee
- Comfort with a catalog whose prices are published before the first call
- Engagement
- Part-time to start; converts when pipeline exceeds founder capacity
- Location
- Capital Region and Hudson Valley; in the room where the buyers are
- Terms
- Pending ratification
- S4OperationsForming
Operations & Finance Lead
Runs the lab's operating rhythm: books, invoicing, compliance filings, the decision log, and the cadence that keeps a founder-led lab honest about its own numbers.
You would own
- Bookkeeping, invoicing, and the New York filings calendar
- The decision log and the weekly operating review
- Vendor and subprocessor inventory, and the counsel and accountant relationships
We look for
- Operations or finance experience inside a small firm that had to be exact
- Fluency with a Notion operating system and a ledger that must reconcile
- A preference for writing the rule down over remembering it
- Engagement
- Part-time to start; converts at the first full quarter of revenue
- Location
- Remote in the United States; Capital Region preferred
- Terms
- Pending ratification
- S5PlatformForming
Platform & Security Engineer
Owns the rail everything ships on: the Cloudflare estate, the access gates, the deploy pipeline, the security baseline, and the backups that are hash-verified rather than assumed.
You would own
- The Workers, D1 and Access estate, and the published quality gate on every deploy
- The security baseline, the incident runbook, and the breach procedure
- Encrypted, verified backups of every estate asset, drilled on a schedule
We look for
- Production experience with edge platforms, identity gates, and CI that fails closed
- A habit of proving a restore instead of trusting a backup
- Comfort being the person who says no to an unauthenticated route
- Engagement
- Contract to start; converts at the first regulated-client deployment
- Location
- Remote in the United States
- Terms
- Pending ratification
- S6ProductForming
Product Lead
Owns what the lab ships as product rather than service: the app factory, the marketplace packs, and the decision about what to build, what to measure, and what to kill.
You would own
- The product roadmap across the apps and the marketplace packs
- Pre-registered hypotheses and the measurement behind each release
- Store listings, pricing proposals for the owner to ratify, and the kill decisions
We look for
- Products shipped to real users, with the retention numbers to discuss
- Comfort writing a hypothesis down before the build and publishing the result after
- A low tolerance for features nobody measured
- Engagement
- Part-time to start; converts at the first paying app or pack
- Location
- Capital Region preferred; remote in the United States considered
- Terms
- Pending ratification
- S7ResearchForming
Applied Research & Evals Lead
Owns the research standard in practice: evals for every agent the lab operates, the pre-registration of every study, and the published findings, including the ones that fail.
You would own
- The eval suites that gate every agent before and after a release
- Pre-registration, baselines, and thresholds on every study the lab publishes
- The research library and the benchmark instruments
We look for
- Experience evaluating language-model systems with methods you can defend
- Statistical literacy and the honesty to publish a null result
- Writing clear enough that a business owner can follow the method
- Engagement
- Contract to start; converts at the second published study
- Location
- Remote in the United States; Capital Region preferred
- Terms
- Pending ratification
- S8BrandForming
Content & Brand Lead
Runs the publishing cadence and keeps the brand exact: one piece every two weeks, consented customer stories, the design-token system, and the brand checks that fail a build.
You would own
- The publishing calendar and the field notes, essays and case studies on it
- Consented customer stories, from consent to two signatures to publication
- The brand standard, the token file, and the lint that enforces both
We look for
- Published writing in a precise, evidence-first register
- Comfort with a design system expressed as tokens and tests, not a mood board
- Patience for the consent process that a real customer story requires
- Engagement
- Part-time to start; converts at the first attributable client conversation from published work
- Location
- Remote in the United States; Capital Region preferred
- Terms
- Pending ratification
- S9HealthForming
Regulated Verticals Lead (Health)
Leads the lab's health work: the hospital pilots, the privacy and safety posture around them, and the clinical and operational relationships that decide whether a pilot is allowed in the building.
You would own
- The hospital pilot relationships and the adoption plan with nursing and supply chain
- The privacy, consent, and safety posture for every health deployment
- The regulatory watch and the written answers to each institution's review board
We look for
- Years inside a hospital or health system in operations, nursing leadership, or informatics
- Working knowledge of health-privacy obligations and how a review board thinks
- A refusal to let a pilot touch patient data before the posture is written down
- Engagement
- Advisor to start; converts at the first signed health-system pilot
- Location
- Capital Region, New York; on site at hospitals
- Terms
- Pending ratification
- S10DeliveryForming
Fixed-Ops Consultant II / Client Success
The second delivery seat for dealerships: carries reviews and pilots alongside the first consultant, then stays with the account as the person who keeps the workflow measured after launch.
You would own
- Reviews and pilots at the second and third dealership accounts
- Post-launch client success: the monthly measurement note and the next workflow
- Renewals, scoped from measured results rather than relationships alone
We look for
- Service-department experience and the patience to coach advisors through a new workflow
- Comfort reading a scorecard and saying plainly what moved and what did not
- A preference for the long relationship over the quick close
- Engagement
- Contract to start; converts at the third active dealership account
- Location
- Capital Region and Hudson Valley; on site at dealerships
- Terms
- Pending ratification
What a listed seat is not
- A seat on this page is a role the lab intends to fill, not an offer of employment. Terms go in writing before anyone commits to anything.
- This is a seat at Blue Horizon Labs, not a position at Anthropic. The lab is a member of the Anthropic Claude Partner Network and nothing more.
- Enrolling in the partner certification path is part of onboarding; any exam result is your own result, and nothing here promises an outcome for you or for the lab's standing in that program.
Terms are not set yet, and this page will not pretend otherwise. Each seat starts as a contract or a part-time engagement and converts at a revenue gate the lab publishes to the person before they commit. Ask in your application; whatever we agree goes in writing first.
The certification cohort
A cohort is forming alongside the seats.
A certification cohort is forming for the January 2027 Claude Partner Network review: a study plan and a project rotation for people who want the certification without taking a seat. Its terms will be published here once compensation is ratified.
What this is not
- A cohort seat is not employment with Blue Horizon Labs, and it is not a position at Anthropic.
- Passing a certification exam is your own result. The cohort provides the study plan, the project rotation, and an @bluehorizonlabs.ai identity for registration, subject to the program's eligibility rule (unverified as of 2026-09-09) — nothing further.
- Nothing here promises that Blue Horizon Labs will reach any particular level of the Claude Partner Network, or that a seat leads to one.
Apply
Send four fields.
No account, no portal, no resume parser. Name the seat you want, or the cohort, and write the rest in your own words.
For businesses
Here as a business, not an applicant?
You are in the right place too. Tell us what you are trying to fix, and we will start with a conversation.