CAPABILITY / 10

Security by design

In production

Access that is scoped, tokens that rotate, data encrypted at rest and sensitive calls on the record.

Updated 15 Nov 2026

01 / PROOF

The measured part

AES-256-GCM
CHECK-enforced encryption in the AI trace store
Source
Rotating
hashed refresh tokens; reuse revokes the family
Source
5 tables
append-only audit trail on the Scheduler
Source

02 / WHAT I DO

In practice

Scope access and re-check identity at use time.
Hash and rotate tokens; revoke on reuse.
Encrypt sensitive records and enforce the shape in the database.
Record sensitive reads in an append-only audit trail.

03 / WHERE I'VE DONE IT

The systems behind it

BadaBhaiAES-256-GCM trace records with a CHECK-enforced ciphertext shape.RVM CAD Job PortalOAuth 2.1 + PKCE and rotating hashed refresh tokens.RVM CAD Job PortalPer-call identity re-check on the read-only AI connector.RVM CAD SchedulerFive-table append-only audit trail and read-only tools.RVM CAD Job PortalConsent-gated recording in restricted Meet spaces, with a 60-minute cap (my statement, 15 Nov 2026).

04 / STACK

What it runs on

PostgreSQL
Shared relational store
See the system
Supabase
Postgres for the public product
See the system
MCP
Read-only connector with 11 tools
See the system
OpenAPI
OpenAPI 3.1 contract, 163 operations
See the system

05 / LIMITS

What this does not claim

Application-layer security. No formal security certification or penetration-test claim.

06 / RELATED CASE STUDIES

All work

Built and testedRVM CAD Job Portal and RVM CAD Scheduler

Both connectors are merged with full test suites. Production activation and the CEO acceptance test had not been verified when this was written (20 Oct 2026).

Doors, not keys: read-only AI access to production data

Two connectors that let an AI assistant answer operational questions from live systems — without a database credential, without write access, and without ever returning a half-right number.

11 read-only tools on the job portal connector

ProductionBadaBhai

Shipped in August 2026; the migration was applied to production with the release. Six AI call sites write to it.

An AI trace store that treats its own data as dirty

Recording every prompt and completion on an AI hiring platform, so bad answers can be investigated — encrypted at rest, erasable per worker, and built so raw text cannot reach the recorder by accident.

6 AI call sites traced

Built and testedRVM CAD Job Portal

Merged with 616 of 616 tests passing at 48c3cab1 (23 Jun 2026). RVM approved describing the behaviour on 15 Nov 2026; production rollout state was not independently verified.

Interviews that schedule, record and file themselves

One interview, end to end in the Job Portal: a Google Meet space and invites, a consent gate before recording, a recording state machine, and Drive filing with metadata. Built in one commit, with 616 of 616 tests passing at that commit.

70 files in the feature commit

07 / QUESTIONS

Short answers

What security work is evidenced?

Application-layer controls: scoped read-only AI access with per-call identity checks, OAuth 2.1 + PKCE with rotating hashed refresh tokens, AES-256-GCM encryption with a database-enforced ciphertext shape, and an append-only audit trail.

Is KP security-certified?

No security certification is claimed. The evidence is the design and tests in the systems above.

Need this on a real system?

Hiring for a role, or scoping a short engagement? The Hire page has the role families and contact; Consulting has the scoped offers.

For hiring managers Consulting