Credentials on the record.
Our team reviews each expert’s employment and work history. Former colleagues can add references.
A standing cell, built to your rubric.
Five to twenty professionals, screened for conflicts against your programs and each expert’s current employer before the first task, and delivered as one unit for the length of an engagement.
Scored answers with the expert’s record attached.
Every answer graded against a published rubric before it ships. First batch: federal acquisition and contracting. Each answer arrives with its record: the expert who wrote it, how their credentials were reviewed, and the rubric it was graded against.
From our CEO · FedScoop Commentary · October 6, 2026
The government is buying AI by the token. It should buy results.
A former DOD official argues for more outcome-driven pricing over usage pricing.
By Savan Kong, CEO and Founder of Your Roster
Read on FedScoopWhere judgment is the product.
A small, trusted pool of operators who did the work inside federal, state and local government. Each one credentialed and qualified problem by problem.
Federal acquisition & contracting
Contracting and procurement officers who wrote the justifications, ran the source selections and lived with the protests.
Defense programs
Program managers from DoD and the services who carried cost, schedule and requirements through milestone decisions.
Federal policy & regulation
Policy analysts and congressional staff who know how a rule is drafted, scored and challenged.
Cleared IT & cyber
IT specialists who ran authorizations, accreditation and cleared-program systems.
State & local procurement
Procurement officers, purchasing agents and finance directors at states, counties and cities who ran competitive bids, cooperative contracts and grant-funded purchases.
State & local policy & regulation
Legislative analysts, agency rule writers and city and county attorneys who know how a statute becomes a rule and an ordinance becomes practice.
Public safety & emergency management
Emergency managers, public safety leaders and grant administrators who ran response plans, mutual aid agreements and disaster recovery funding.
Health & human services
Medicaid, public health and social services staff who ran eligibility, program integrity and state plan rules.
From your model’s answers to reinforcement learning (RL) training data.
Credentialed government experts, reviewed by our team, grade your model’s answers and write out the reasoning it’s missing. Every grade and every trace becomes training data for RL.
- 01 · Scope
A 30-minute call
Tell us the domain, the kind of work and the rough scale. Nothing to sign while it’s in build.
Engagement BriefReady for a CallDomainSource selection, FAR 15.3Kind of workReasoning traces and scored answersModel accessAPIFirst deliveryOne package, binned by outcome - 02 · Grade
Experts score your model’s answers
Credentialed experts grade each answer your model gives against the rubric, one dimension at a time. Each grade becomes a reward signal your model can learn from.
Answers Graded by AreaIllustrativeSource selection31Describing agency needs22Contract types14Bid protests9 - 03 · Capture
Experts write the reasoning
Where an answer falls short, the expert writes out the right steps, each with the rule it rests on. That reasoning trace shows your model what good work looks like.
Reasoning Trace · 14:32 on Problem1No relevant record means no favorable or unfavorable rating.FAR 15.305(a)(2)(iv)2Check how the solicitation says a neutral rating is treated.Section M3Record the reasoning in the source selection decision.FAR 15.308 - 04 · Deliver
A standing cell or a sealed dataset
Grades and reasoning traces arrive as one sealed package, ready for your RL pipeline. Every record is hashed and comes only from experts qualified for the problem.
See where your model is confidently wrong before you spend an expert hour.
A model that is wrong and unsure already hedges. A model that is wrong and confident gives a contracting officer a wrong answer in the tone of a right one. Move the line and watch the queue reorder.
Sample agreement across five runs of the same question.
An offeror has no record of relevant past performance. How should the evaluation team rate it?
Model: “Rate it Marginal, since the offeror cannot show relevant experience.” Agreement 5 of 5. Confidently wrong.
- Expert
- K-17 · contracting officer · reviewed
- Time
- 14:32
- Saw model answer
- false
- Record
- sha256: 7c1e94b0a52fd8316e0c4a9b27d53f81e6a0c9d42b7f15e38a60d2c9e41b7f5a
Models learn the outcome. Experts carry the reasoning.
Every trace records each step, what the step rests on, the answer, and how long it took. The expert never sees the reference answer.
Request a Sample PackageThree rules for how we score.
Aegis Eval, the scoring engine underneath Your Roster, follows three rules and is tested against all three. A change that breaks one doesn’t ship.
No magic number
Each dimension gets its own result. We never combine them into one trust score, because how much each one matters depends on your mission, and that’s your call.
Unknown is a valid result
Anything we didn’t test is marked NOT EVALUATED. A run that couldn’t reach a judgment never counts as a pass.
Evidence over claims
Every score links back to the question, the source material, the model’s output, the expert’s reasoning and each grading decision, plus a sha256 fingerprint of the stored record, so any later change shows.
Why work with us? Other expert-data vendors say “passed our assessment.” We show you the credentials behind every answer.
Built for the buyers who can’t afford to guess.
AI labs
Reasoning traces and scored answers on the government problems your model gets wrong, binned so the valuable data is never sold at the bulk rate.
Federal integrators
Cleared and former-federal expertise for programs where an identity credential alone doesn’t establish judgment quality.
Program offices
Evidence for a fielding decision: a model evaluated against your mission profile, with every result traced to the record behind it.
Teach your models the judgment they’re missing.
Design partners help shape the rubric and the makeup of their cell while both are still open.
- A 30-minute call
- A failure map for your model
- A sample package with each expert’s record
- Nothing to sign while it’s in build