Back to Projects

Sindri

Applied AI Engineer Experimental Access on request

Updated: 2026-04-04

Capability-synthesis service that builds wrapper artifacts, runs verifier-guided review loops, and enforces promotion gates before any adoption.

Impact: Turns speculative automation into bounded, reviewable capability proposals with explicit safety and graduation criteria.

What I built

  • Runs phased trust calibration with pilot wrapper-lane corpora, disagreement analysis, and regression guardrails.
  • Produces structured review bundles, patchability assessments, and promotion-gate outputs for operator decisions.
  • Phase 11 extends the model into adjacent-lane transfer proving while keeping unsafe promotion at zero.

Proof: Run `make eval_pilot_wrapper`, `make eval`, or `scripts/run_phase11_adjacent_lane_transfer.sh` for reproducible trust-lane evidence.

PythonFastAPIPytestJSON