Synapse Insight
Demo project — an illustrative engagement showing how Quadric-ai scopes, builds and measures work of this type. Named client case studies are published as clients approve them.
An AI analytics engine that reads unstructured market and filing data, then surfaces forecasts, anomalies and decision-ready briefings for analysts.
- Client
- Synapse Capital
- Year
- 2026
- Duration
- 6 months
- Team
- 3 ML engineers, 2 data engineers
The challenge
- Analysts spent most of the week reading documents rather than deciding.
- Signals were buried across filings, transcripts and internal notes.
- Any model output had to be explainable and source-linked.
Services
- Applied AI
- Data Platform
- MLOps
Our approach
01
Retrieval with citations
A vector + keyword hybrid index grounds every generated insight in a source passage the analyst can open.
02
Anomaly detection
Time-series models flag divergence from peer baselines and push ranked alerts into the daily briefing.
03
Evaluation harness
A golden dataset and automated evals gate every model or prompt change before it reaches production.
Results
-70%
Research time per brief
94%
Citation accuracy
3.2M
Documents indexed
12
Analyst seats live
Technologies used
- Python
- PyTorch
- LangChain
- OpenAI
- dbt
“The difference is trust. Every insight links back to the source line, so our analysts adopted it within a fortnight.”
Have a project like Synapse Insight?
Tell us what you are building and we'll come back with scope, plan and estimate.
Start your project