Selected project / NewsAI / MarketGuide
Turn financial broadcasts into researchable events.
NewsAI connects financial video and audio with speaker-attributed transcripts, structured LLM analysis and historical price data.
01 / The challenge
NewsAI / MarketGuide
Useful information is spread across long broadcasts, interviews and press conferences. Researchers need to find who said what, inspect the original context and explore the market response.
The engineering approach
CUDA-accelerated diarization segments speakers. Transcription and LLM workflows organise topics and events, while price-data alignment supports historical research and backtesting. Source timestamps keep the analysis connected to the original media.
02 / System architecture
Turn financial broadcasts into researchable events.
CUDA-accelerated diarization segments speakers. Transcription and LLM workflows organise topics and events, while price-data alignment supports historical research and backtesting. Source timestamps keep the analysis connected to the original media.
03 / From input to useful output
What the system connects
Ingest media
CUDA speaker diarization
Separate speakers
Time-aligned transcription
Extract events
Structured LLM pipeline
Align prices & backtest
Price data & backtesting
Contact
Working on a similar challenge?
Tell us about the workflow, the constraints and the result you need.