Operations Intelligence
Operational Analytics & Decision Support
Operations Intelligence is a deterministic operational analytics platform that helps businesses understand performance, identify bottlenecks, detect anomalies and surface evidence-backed improvement opportunities. The LLM layer interprets validated evidence and produces narrative explanations without performing calculations.
Phynapse research & engineering project · Synthetic data
The business problem
Operational data (throughput, cycle times, queues, SLAs, vendor delays) is usually spread across systems and reports. Leaders see symptoms but struggle to locate the constraint.
Narrative AI summaries without a deterministic foundation can point teams at the wrong problem.
What we engineered
Structured operational datasets are processed by a set of deterministic analyzers. Results are assembled into evidence packages, and a language model may then explain that evidence in plain language. Its output is treated as untrusted and re-validated before use.
What the system does
Bottleneck analysis
Constraint ranking with explainable scores.
Cycle-time decomposition
Queue versus processing time, by process and department.
SLA and backlog analysis
Compliance, breach exposure and accumulation of open work.
Anomaly detection
Unusual operational patterns surfaced for review.
Vendor and department views
Third-party delay contribution and per-department scorecards.
Natural-language questions
Operational questions answered from validated evidence.
Design decisions
- 01
Twelve catalogued analysis types, each implemented deterministically.
- 02
User-supplied identifiers are resolved through a fixed registry; data paths cannot escape the data directory.
- 03
Secrets live only in environment variables and are never exposed to the frontend.
- 04
Structured error handling ensures stack traces are never returned to clients.
How information moves through the system
- 1
Operational datasets
Throughput, orders, SLAs, vendors, departments.
- 2
Analyzers
Deterministic modules for each analysis type.
- 3
Evidence service
Top contributors and findings packaged as evidence.
- 4
AI interpretation
Narrative explanation of validated evidence.
- 5
Dashboard & answers
Decision support for operations leaders.
Built with
- Python 3.11+
- FastAPI
- Pydantic
- Pandas
- React + Vite
- TypeScript
- pytest
- Docker
As documented in the project repository.
What this project is, and is not
- Engineered portfolio system built on synthetic operational datasets.
- Narrative features require a configured OpenAI-compatible model endpoint.
- Analyses depend on the structure of the supplied datasets; new sources require mapping.
Review the engineering in the open.
The full source, documentation and tests for Operations Intelligence are available on GitHub.