GADEAPPLIED
Discuss a project

Case studies

Business context.
Engineering depth.

From expert-led AI Agent evaluation to enterprise privacy operations, we turn domain knowledge and business responsibilities into working systems.

Explore the projects
01

AI Agent · Evaluation engineering

A well-known enterprise Client anonymized

From expert tasks to executable agent evaluations.

Expert task annotation, reward scripts, and rollout engineering for scientific AI agents, informed by the public Terminal-Bench-Science benchmark.

Business challenge

Scientific expertise had to become work an agent could execute and a team could evaluate. Instructions, datasets, runtime environments, output artifacts, and scoring rules needed to agree without exposing the answers to the agent.

Public reference: Terminal-Bench-Science

What we delivered

  • Coordinated expert task annotation and translated domain requirements into task instructions, output contracts, and reward scripts.
  • Packaged tasks into containerized environments, connected agent rollouts to artifact collection and evaluation, and separated public inputs from hidden grading assets.
  • Built validation tools for task structure, output and grader alignment, scoring behavior, and detailed evaluation records so experts and engineers could inspect the same evidence.
Delivery architecture

Expert knowledge becomes executable evaluation.

SIMPLIFIED VIEW
  1. Expert tasksInstructions, data & output contracts
  2. Task environmentContainers & runtime constraints
  3. Agent rolloutTask execution & run records
  4. Output artifactsFiles, code & research outputs
  5. Reward evaluationChecks, scores & detailed logs
Simplified delivery architecture. Hidden grading assets are available to evaluation and kept separate from the agent environment.
Delivered outcome

A system built for the work.

Executable task packages, reward scripts, rollout workflows, and traceable evaluation records: a practical foundation for testing agent behavior and refining tasks with domain experts.

FDE in practice

Business and engineering, together.

Our engineers worked through the scientific workflow with experts, resolving the details of inputs, deliverables, runtime constraints, and scoring together. Domain knowledge became running software and inspectable evidence.

  • Harbor task packaging
  • Container environments
  • Rollout orchestration
  • Reward evaluation
  • Artifact validation
02

Custom software · Privacy operations

A well-known enterprise Client anonymized

One platform for a cross-team privacy workflow.

A custom Data Subject Request platform connecting consumer and employee requests with business, privacy, and IT teams.

Business challenge

A request can cross business or HR reviewers, privacy specialists, and IT before a response is ready. The enterprise needed a shared workflow for ownership, deadlines, evidence, and controlled access to personal data.

What we delivered

  • Developed request registration and a staged workflow for business or HR review, privacy review, IT processing, result verification, and response records, with deadlines, assignments, and evidence tracking.
  • Integrated enterprise OIDC single sign-on with server-side permissions based on role and business scope. Added field encryption, default masking, and authorized, audited access to sensitive details.
  • Connected workflow transitions and audit records within the same database transaction, with private file storage and an outbox for queued team notifications.
Business workflow & architecture

Built around each team’s responsibilities.

SIMPLIFIED VIEW
  1. RegisterConsumer or employee request
  2. Initial reviewBusiness or HR team
  3. Privacy reviewPrivacy team or DPO
  4. ProcessIT team takes action
  5. VerifyReview the result & evidence
  6. FeedbackRecord the response & evidence

Across every stage Ownership · Deadlines · Evidence · Access · Audit history

IdentityEnterprise SSOOIDC sign-in
Application boundaryWorkspace & API
  • Role & business-scope authorization
  • Encrypted fields & masked views
  • Authorized, audited data reveal
  • Workflow & audit databaseTransitions and audit records in one transaction
  • Private file storageAccess-controlled request evidence
  • Notification outboxQueued team notifications
Simplified business and system architecture. Request routing follows consumer or employee responsibilities; personal-data controls sit within the application boundary.
Delivered outcome

A system built for the work.

A custom platform that brings request handling, team responsibilities, deadlines, evidence, access controls, and audit history into one coherent workflow.

FDE in practice

Business and engineering, together.

We translated the responsibilities of business, HR, privacy, and IT into workflow stages and access boundaries. The software reflects how the teams handle requests, including handoffs and the evidence needed at each stage.

  • OIDC SSO
  • Server-side authorization
  • PII controls
  • Transactional audit
  • Private storage
  • Notification outbox
03

Enterprise software · Continued delivery

A well-known enterprise Client anonymized

A product carried from definition into live operation.

A two-year enterprise software engagement spanning product definition, engineering, deployment, and continued operation.

Business challenge

The enterprise needed an engineering partner to define a business-critical software product and remain responsible as it moved through development and into everyday operation.

What we delivered

  • Defined the product and delivered the engineering work through deployment under a two-year enterprise services engagement.
  • Continued operating and improving the product after launch, maintaining responsibility across the delivery lifecycle.
  1. Product definition
  2. Engineering
  3. Deployment
  4. Continued operation
Delivered outcome

Responsibility beyond launch.

The product entered live operation, with GADE continuing to maintain and improve it beyond launch.

FDE in practice

Business and engineering, together.

Delivery responsibility continued from the first product decisions through deployment and ongoing operation, keeping engineering connected to the business throughout the engagement.

  • Product definition
  • Software engineering
  • Deployment
  • Ongoing operation

Your business challenge

What does your team
need to make work?

Bring us an Agent idea, a software requirement, or a workflow that needs an engineering team.

Discuss your business challenge