Sycra
SYCRA is an AI-powered platform designed for sales and telemarketing teams. It runs in the background during live calls, captures and processes audio in near real time, generates transcriptions, extracts structured customer information, automatically updates CRM fields, and supports post-call proposal generation.
Visit Landing Page
My Contributions
- As Technical Co-Founder, I am responsible for the end-to-end technical design and implementation of the platform, working across the full stack - from backend architecture, audio processing and AI pipelines to infrastructure, integrations, and the user-facing application.
Tech Stack
- Python & FastAPI - Backend APIs and core application logic
- PostgreSQL - Persistent storage for calls, transcripts, extracted data, and system state
- Redis - Asynchronous job queues and communication between processing components
- Distil-Whisper - GPU-accelerated speech-to-text
- spaCy & LLMs - Entity detection, information extraction, validation, and reasoning
- MinIO - Object storage for audio and generated files
- Docker - Containerization and service orchestration
- GitHub Actions & Pytest - CI/CD and automated testing
Features
- Automatic live call detection
- Background audio capture
- Overlapping audio segmentation
- Voice Activity Detection and audio enhancement
- Real-time speech-to-text
- AI-powered information extraction
- Automatic CRM synchronization
- Live transcription interface
- Call session and audio chunk tracking
- Persistent audio storage
- Proposal generation
- Concurrent GPU processing
Project Status
Discontinued - Development of SYCRA has been discontinued following changes in the market and the broader competitive landscape. Although it does not progress to full commercial deployment, SYCRA remains a valuable engineering project and a significant part of my experience building a complex product from its initial technical design toward a production-ready system.
