BaseBuildStack: Privacy-First AI Technical Interview Preparation Built on Open Weights

Описание и применение

BaseBuildStack – микросервис, использующий локальные модели открытого веса для подготовки к техническим интервью, обеспечивая низкую стоимость и защиту данных.

Задача

Commercial interview prep platforms are expensive and pose data privacy risks by uploading sensitive candidate data to third‑party LLM services.

Результат

The platform delivers adaptive interview practice at a fraction of the cost while keeping all candidate data on the local machine, enabling offline use and future‑proofing against API changes.

Рабочий процесс

• 1. Candidate registers via Google/GitHub OAuth or email.
• 2. The app extracts candidate tech stack and resume claims locally.
• 3. Open‑weight models (gemma2:9b, llama3) generate interview questions and hints through Ollama.
• 4. Generated code tests are executed in a deterministic Node.js sandbox to validate answers.
• 5. Responses are parsed and validated against Zod schemas for type safety.
• 6. Adaptive quizzes and spaced‑repetition cards are presented to the user.
• 7. User can review performance and repeat drills offline.

Доказательства

BaseBuildStack delivers an adaptive, high-speed interview practice suite that runs on open-source AI

Источник

Local Open-Weight Inference (Ollama) … interfaces with local open-weight models (gemma2:9b and llama3) via Ollama

Источник

Deterministic Assertion Sandbox … executes generated test suites deterministically against candidate code

Источник

Zod Boundary Schemas … parsed and strictly validated against Zod schemas (LlmValidationError) to guarantee rigid runtime types

Источник