https://leetspeak.live/

Mock interview practice that actually interviews you back.

LeetSpeak gives you realistic behavioral, technical, and system design interview rounds with AI-powered feedback. Upload your resume, pick a scenario, talk through your answer, and get graded on a real rubric — not a generic "good job."

Practice interviews across three categories:

• Behavioral — "Tell me about yourself," leadership stories, conflict resolution
• Technical — Coding problems with a live Monaco editor, LeetCode integration, BrainStellar-backed quant problems, diagram whiteboard (tldraw)
• System design — Architecture discussions with follow-up questions

AI grading and coaching:

• Interviews are graded by LLM (OpenAI or Gemini) against scenario-specific rubrics
• Resume text is extracted and fed into grading context so feedback is personalized
• A separate AI Coach chat lets you ask questions, get tips, or debrief between rounds

Track your progress:

• Dashboard shows interview history, scores, streaks, and improvement areas
• Profile page displays mastery across different tracks and difficulty levels
• Review pages break down each attempt: score, rubric breakdown, strengths, areas to improve

• Framework: Next.js 16 (App Router), React 19, TypeScript
• Styling: Tailwind CSS v4, DaisyUI, shadcn/ui primitives
• Auth: NextAuth.js with Google OAuth (JWT sessions)
• Database: MongoDB Atlas — interview persistence, user profiles, metrics
• LLM: Provider abstraction supporting OpenAI-compatible endpoints and Google Gemini
• Code editor: Monaco Editor (VS Code's editor)
• Whiteboard: tldraw for diagram/system design sketches
• Voice/Avatar: HeyGen LiveAvatar + ElevenLabs integration
• Document parsing: pdf-text-extract + mammoth (PDF, DOCX resume upload)
• Deployment: Cloudflare Workers via OpenNext

npm install
npm run dev

Copy .dev.vars.example to .dev.vars and fill in your keys. The app degrades gracefully when integrations are missing — auth, LLM, and DB features just won't be available.

app/                    → Next.js App Router pages and API routes
  (shell)/              → Main app layout (sidebar nav, auth gating)
    practice/           → Scenario picker + live interview sessions
    profile/            → User profile, resume upload, mastery stats
    coach/              → AI coach chat
  api/
    interview/          → Start, end, transcript, grading endpoints
    session/            → LiveAvatar/ElevenLabs session management
    auth/               → NextAuth handler
components/app/         → Feature components (practice session, coach, settings)
components/ui/          → Reusable UI primitives (button, card, badge, etc.)
data/                   → Scenario bank, LeetCode problems, fixtures
lib/                    → Auth config, MongoDB client, LLM router, doc extraction
lib/integrations/       → LLM provider implementations (OpenAI, Gemini)

All LLM calls route through lib/integrations/llm.ts for centralized provider selection and fallback handling.

Built at HackDartmouth XI.

Quant problem prompts in the practice flow are sourced from the BrainStellar repository by GitHub user rudradesai200.