Technical diligence room

InterviewCoach

Investor review flow for the Android product, AI orchestration layer, release system, observability, load evidence, and cost controls.

Download release APK Open deep dive
Release artifact Signed Android APK and Play handoff packet
Load baseline 111723 requests, 186.24 req/s, p95 800 ms
Trace proof 17 runs, 5 traces, 1 thread in smoke proof
Delivery guard 8/8 public checks, 5/5 gates passing

Review path.

Start with the public proof, then use the protected deep dive for source files, run outputs, and limits.

  1. 01
    Product and mobile release

    Inspect the APK, Play screenshots, metadata packet, and deterministic Android release gates.

  2. 02
    AI orchestration and tracing

    Review Python orchestration, retrieval boundaries, report generation, and the LangSmith trace smoke.

  3. 03
    Scale and cost control

    Separate deterministic load testing from tiny live-model sampling so throughput and spend are governed independently.

  4. 04
    Honest diligence boundary

    Keep iOS, AWS execution, and live-model stress tests as gated next proofs until those runs exist.

Evidence now.

Current claims are tied to code, generated artifacts, smoke reports, or live public-safe endpoints.

Agentic orchestration

Interview generation, answer coaching, report creation, retrieval context, and traceable orchestration boundaries are implemented in the backend.

Mobile delivery loop

Android release gates cover deterministic states across setup, chat, post-answer, completed sessions, and review/history flows.

Cost governance

Load testing runs on deterministic runtime. The first live-model trace sample is capped, small, and separate from stress testing.

Runtime scale path

Codex Spark remains the alpha proof path; OpenAI API is now a config-switchable serving path for future Google Play traffic.

Trace evidence

LangSmith tracing is proven by deterministic and live-model public-scenario smokes, separate from stress tests so traces stay readable.

Delivery discipline

Repo/release evidence shows 94 commits since May 18, 50 release/play/site-related commits, and current launch gates passing.

Protected boundary

The deep dive is gated separately: public pages stay open, protected evidence is Basic Auth scoped, and secret/raw-data classes stay out of git and static pages.

Investor criteria coverage.

The protected answer map links each role requirement to current proof and explicitly marks roadmap items.

ProvenAI/RAG, Python orchestration, Android release gates, deterministic and live LangSmith trace smokes
Measured150-user deterministic mobile-flow load, 111723 requests, p95 800 ms, zero-failure baseline
ControlledModel spend separation, runtime provider switch, $0 load spend, $3 live-test cap, protected room
Guarded8/8 readiness surfaces live, 5/5 current gates passing, protected deep dive excludes raw CV and credentials
Gated nextPaid API scale sampling, AWS execution, Fastlane upload, iOS/XCUITest pattern

Live surfaces.

Public-safe endpoints stay open; sensitive traces, raw data, and deeper test artifacts belong behind the protected deep dive.

Ops pagePublic-safe runtime and operator summary. Test evidenceReadable job-match and interview simulation proof. Ops summaryBackend totals and dependency readiness. RuntimeAgent runtime, model path, tracing, and surface links. Metrics summaryCurated counters and runtime latency shape.