AI Copilot Overview

Amigo's AI copilot is the engine behind every interview session. It listens to the conversation in real time, understands the question being asked, and surfaces tailored answers — all before you've finished thinking.

How It Works

When you start a session, Amigo runs a pre-flight system readiness check — verifying stealth mode, audio permissions, and network connectivity — then captures the audio stream from your interview call and passes it through a multi-stage pipeline:

1. Capture — The desktop app taps into your system audio so it hears both sides of the call. A separate candidate microphone stream runs in the background, allowing Amigo to distinguish your voice from the interviewer's and avoid triggering suggestions from your own speech.
2. Transcribe — Speech is streamed to a low-latency cloud transcription service via the secure native backend — no API keys or audio data ever pass through the browser. Interim transcripts appear progressively as the interviewer speaks, with the final transcript delivered the moment they pause.
3. Filter — A dedicated filler-rejection layer (Layer 0) strips out elongated sounds ("umm", "uhh"), conversational fillers ("okay", "so yeah"), and candidate speech before anything reaches the AI — so only substantive interviewer questions trigger a response.
4. Classify — Amigo's NLP engine uses grammar-aware pattern matching to determine whether the question is coding-related or conversational, then selects the best-fit response strategy. A theory-frame guard ensures that explanation-style questions ("explain how a HashMap works") are not misclassified as coding tasks. Short follow-up questions ("and what about edge cases?") inherit the intent from the prior question using dialog-context analysis.
5. Generate — A personalised answer is drafted using your resume, target role, and the active system prompt. The model tier is selected automatically: coding questions use a more capable model, while conversational questions use a fast-response model. In Coding Test mode, the most capable model is always used.
6. Display — The suggestion streams into the Amigo overlay token by token — visible only to you, invisible to screen share — so you can start reading before the full answer has finished generating.

Session Formats

When you set up a Copilot Profile on the web app, you choose a session format that tells Amigo how to handle your interview:

Response Length

Control how much detail Amigo includes in each suggestion. Tap the response-length pill in the overlay toolbar to cycle between three levels:

Your preferred response length is saved per profile and synced automatically when you start a session. You can also change it on the fly during the interview without interrupting the conversation.

Personalisation

Amigo tailors every answer using three layers of context:

• Resume — Projects, companies, and skills are extracted so answers cite real evidence from your history.
• Target Role — The job title and description you provide shifts vocabulary, seniority signals, and domain focus.
• Copilot Profiles — Custom instructions let you enforce tone, length, session format, or any other constraint (e.g. "always end with a question for the interviewer"). Each profile can specify its own response length and session format.

Smart Filtering

Not every sound in a call deserves an AI response. Amigo uses a multi-layer filtering system to ensure suggestions are triggered only by genuine interviewer questions:

• Candidate speech detection — Amigo captures your microphone separately and automatically ignores your own voice, so your answers don't trigger new suggestions.
• Filler rejection — Over 100 common filler phrases ("let me think", "you know what I mean", "basically") and elongated vocalisations ("ummmm", "uhhh") are filtered out before classification.
• Sentence hygiene — For voice transcripts, preamble and filler sentences are stripped so only the substantive question reaches the AI.
• Dialog-context awareness — Short follow-up questions that lack clear signal on their own inherit the intent of the previous interviewer question, so Amigo stays on topic across multi-turn exchanges.

Latency & Accuracy

Amigo is optimised for speed. In typical conditions on a stable internet connection:

Audio is routed through the secure native backend to a cloud transcription service — no API keys or audio data ever pass through the browser. Interim transcripts stream progressively so Amigo begins processing before the speaker finishes. Accuracy depends on microphone quality and background noise; a wired headset or noise-cancelling earbuds are recommended for best results.

System Readiness Check

Before every session, Amigo runs an automatic pre-flight check to make sure your system is ready. This includes verifying that stealth mode (window exclusion) is active, the overlay is configured to stay on top, audio permissions are granted, and network connectivity is available. If anything needs attention, Amigo will prompt you to fix it before the session begins — so you never start an interview with a misconfigured setup.

Interview Briefs

Briefs are rich-text reference notes you can prepare before an interview and access from the Amigo desktop app during your session. Use them to jot down key talking points, company research, or answers to questions you expect.

• Create and manage briefs from the My Briefs page on the web app.
• Your brief is available automatically when you sign in on the desktop app.
• During the interview, open the Brief Panel to view your notes alongside the AI suggestions.
• Briefs auto-save every 5 seconds, so your edits are never lost.

Resumes & Job Descriptions

Upload your resume and job descriptions from the All My Docs page. These documents are linked to your Copilot Profile and used by the AI to personalise every suggestion with your real experience and the specific role you're targeting.

• Supported formats: PDF, Word (.doc, .docx), and plain text (.txt).
• Maximum file size: 10 MB per document.
• You can also paste text directly if you prefer not to upload a file.
• Link documents to a profile from the Copilot Profiles page — each profile can have its own resume and JD.

Screenshot Capture

During a session, you can capture a screenshot and send it directly to the AI for analysis. This is useful for coding interviews where the problem statement is displayed on screen, or when you want Amigo to read a diagram or table the interviewer is sharing.

• Trigger: Press Ctrl+Shift+S (Windows) or Cmd+Shift+S (macOS).
• Select area: A capture overlay appears — drag to select the region you want to capture.
• Auto-send: The screenshot is automatically sent to the AI, which analyses the content and generates a response using a screenshot-aware prompt.

Keyboard Shortcuts

Amigo supports global keyboard shortcuts that work even when the app is not focused. All shortcuts are customisable from Settings inside the desktop app.

Token Usage

Each generated suggestion consumes tokens from your account balance. Longer answers and complex coding explanations use more tokens than short conversational replies. Choosing the Concise response length helps conserve tokens when speed matters more than depth. You can monitor your consumption on the Sessions page after each interview.