"Your AI that knows who you're talking to before you do."

CORTIS

Context Intelligence Platform

Real-time, on-device AI for high-stakes conversations. Ingests your calendar, email, contacts, and CRM — delivers the right context at the right moment through any device.

meeting notes post-call analysis passive transcription → live runtime intelligence

Solo Founder Build ex-AWS / DoorDash / Cruise

Active Development — UI Design Complete

// the problem

Every Professional Goes In Blind

Meeting prep takes hours. Interview research is scattered across 10 tabs. Context about who you're talking to lives in your head or nowhere. You know you have relevant information somewhere — but you can't surface it in the moment that matters.

CORTIS closes that gap. It's a context intelligence platform that ingests your life context — calendar, email, contacts, LinkedIn, CRM — and delivers actionable intelligence before and during every conversation. Not after. Not as a summary. In real time.

Built by a solo founder with experience shipping real-time systems at Cruise (autonomous vehicles, millisecond decisions), infrastructure at AWS, and personalization at DoorDash. Currently experiencing the exact problem CORTIS solves — interviewing while building.

// core intelligence

Four Pillars

Pre-fetch, don't react. The latency secret is doing work before you need it.

🎙

Live Conversation Processing

On-device speech processing. Understands context, intent, and nuance as the conversation unfolds — no cloud round-trip.

🔍

Context Surfacing

Proactively surfaces relevant history, documents, and data points. Pre-fetched for known events (<500ms), live for spontaneous (<2s).

💬

Response Intelligence

Suggests talking points grounded in your actual knowledge base. Delivered via phone, earbuds, or smart glasses.

🧠

Personal Knowledge Graph

Builds a compounding graph from every interaction. Hot cache of top 200-500 contacts with pre-built profiles.

// use cases

Same Platform, Many Verticals

All use cases share the same architecture: context ingestion, intelligence synthesis, briefing generation, post-analysis. Only the data sources and templates change.

Beachhead — Build First

Interview Preparation

Auto-detect interviews from calendar. Generate company briefings, interviewer intel, talking points. Real-time assist during calls. Post-interview analysis and pipeline tracking.

Month 4-5

Sales Call Assist

CRM integration, pre-call briefings with account history, real-time objection handling, auto-generated follow-ups.

Expansion

Networking & Events

Scan attendee lists, pre-build profiles. Quick lookup via phone or smart glasses during events. Post-event follow-up generation.

Expansion

Executive Meeting Prep

Universal calendar intelligence. Who's attending, last discussion summary, agenda, talking points. Works for any meeting.

// architecture

Six-Layer Stack

From data ingestion to sub-200ms delivery.

Data Ingestion

Manual input via Expo app (MVP). Google Calendar webhooks + batch polling planned for future.

Personal Knowledge Graph

Supabase pgvector for embeddings + relational tables for contacts, companies, and relationships. Thin abstraction for future Pinecone swap.

Context Fusion Engine

LangGraph orchestration: fetch context → rank → sanitize → generate. Input sanitization for LLM injection protection.

LLM Reasoning

Claude Haiku for speed, Sonnet for depth. Retry 2x with backoff, graceful fallback to raw context. Full prompt/response logging.

Delivery Layer

Expo universal app (iOS, Android, Web). Push notifications + in-app briefing cards. Deepgram cloud transcription for audio.

On-Device Intelligence (Planned)

Expo-modules wrapping CoreML (iOS) / NNAPI (Android) for real-time on-device inference. Speaker diarization and Meta glasses SDK deferred.

// prerequisite learning

Active Coursework

Four DeepLearning.AI courses mapping directly to CORTIS's technical architecture.

COURSE 01

Introduction to On-Device AI

→ Edge inference layer

Mobile inference and hardware constraints for running AI on phones and smart glasses without cloud round-trips.

COURSE 02

Building AI Voice Agents for Production

→ Real-time conversation pipeline

Latency management, voice-based AI orchestration, and production challenges of real-time audio processing.

COURSE 03

Quantization Fundamentals with Hugging Face

→ Model compression layer

Compress models to run on phones and glasses without lag. Practical skills for constrained hardware.

COURSE 04

AI Agents in LangGraph

→ Autonomous orchestration backbone

Multi-step agent workflows: read calendar, detect event, pull data, fetch profiles, generate briefing, deliver.

// built for

Anyone Whose Conversations Have Consequences

🚀

Founders & Executives

Board meetings, investor calls, strategy

💼

Sales Teams

Client calls, deal reviews, competitive intel

🎯

Consultants

Client engagements, cross-project context

📋

Product Leaders

Stakeholder alignment, roadmap decisions

// ui design — 19 screens designed

Interface Design

Premium dark aesthetic. 19 screens covering core flows, error states, loading states, and edge cases. Expo universal app targeting iOS, Android, and Web from a single codebase.

// design system highlights

Instrument Serif + DM Sans Amber/Gold Accent on Deep Dark 19 Screen Designs Error States Loading Skeletons Pipeline Animation Share Sheet Pull to Refresh

App Store — Coming Soon Play Store — Coming Soon Web App — Coming Soon

// development status

Build Progress

4-week MVP roadmap. Currently in the UI design phase with architecture decisions finalized.

✓

Architecture Review

Engineering + CEO review complete. Vector store, orchestration, ingestion, security, and observability decisions finalized.

✓

UI/UX Design — 19 Screens

Core flows, error states, loading states, empty states, settings, share sheet, pull-to-refresh. All review conditions passed.

✓

Task Breakdown — 19 Tasks

4 milestones: Foundation, Intelligence Pipeline, Audio + Knowledge, Deploy + Dogfood. Tracked in GitHub Projects.

◯

Milestone 1: Foundation

Monorepo setup, Supabase schema + RLS, FastAPI skeleton, Expo auth flow.

◯

Milestone 2: Intelligence Pipeline

Input parser, knowledge store, Proxycurl enrichment, LangGraph briefing pipeline, Expo UI.

◯

Milestone 3: Audio + Knowledge

Deepgram transcription, transcript analysis, audio recording, knowledge graph feedback loop.

◯

Milestone 4: Deploy + Dogfood

Railway/Render deployment, Expo EAS builds, test suites, dogfood for real interviews.

// tech stack

Under the Hood

Python (FastAPI) Supabase (Auth + pgvector + RLS) Anthropic Claude (Haiku + Sonnet) Deepgram Cloud LangGraph Expo SDK 54 (iOS + Android + Web) Proxycurl (LinkedIn) structlog + Sentry Railway / Render