Skip to content

Latest commit

 

History

11 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

AI-Powered Customer Operations Command Center

This repository analyzes customer support ticket data, runs an agentic weekly ops pipeline, and generates a structured Ops Intelligence Brief with the top improvement opportunities, trend changes, and recommended actions.

What this repo includes

Main capabilities

  • Cleans messy support ticket data with explicit handling of invalid and missing values
  • Computes KPI summaries, weekly trends, team/channel/category performance, and outliers
  • Analyzes customer text for sentiment, frustration, and topic signals
  • Runs a LangGraph pipeline to synthesize findings into opportunities and a weekly brief
  • Presents results in a 5-tab Streamlit dashboard

App overview

The app includes five views:

  • Dashboard — KPIs, performance comparisons, and trends
  • NLP Insights — sentiment, frustration, and text patterns
  • Trends — week-over-week movements and anomaly review
  • Opportunities — prioritized improvement actions with impact estimates
  • Weekly Brief — a generated markdown summary for leadership review

Architecture

The project has two main layers:

1. Deterministic analytics

Core modules:

2. Agentic synthesis

Core modules:

Pipeline flow:

ingest → data quality → parallel analysis (trends / anomalies / NLP) → opportunity scoring → report generation → executive insights

Tech stack

Component Implementation
Language Python 3.11+
Frontend Streamlit
Agent Framework LangGraph + LangChain
LLM Groq (meta-llama/llama-4-scout-17b-16e-instruct)
Data Processing pandas, numpy
Statistics / ML scipy, scikit-learn
NLP vaderSentiment + sklearn-based text workflows
Visualization Plotly
Testing pytest

Key findings from the sample dataset

Top modeled opportunities identified in the analysis:

Rank Opportunity Estimated Annual Impact
1 Email to chat deflection $215,731
2 Abandonment reduction $199,861
3 Chatbot escalation reduction $192,408
4 Proactive CSAT recovery $184,176
5 Vendor B quality improvement $49,561
Total modeled impact $841,737

A few representative findings:

  • Email is the largest and one of the least efficient channels
  • Refund and order-status tickets drive a large share of workload
  • Vendor B materially underperforms on speed and CSAT
  • The chatbot is extremely cheap and fast, but escalates too often in the wrong categories
  • Volume rises across the complete weeks without clear efficiency gains

Repository structure

G/
├── data/                       # Input dataset
├── docs/                       # Assignment materials
├── notebooks/                  # EDA notebook
├── output/                     # Generated artifacts
├── slides/                     # Slide content and roadmap notes
├── src/
│   ├── analytics.py
│   ├── config.py
│   ├── data_cleaning.py
│   ├── nlp_analysis.py
│   ├── visualizations.py
│   ├── agent/
│   └── app/
├── tests/
├── requirements.txt
└── README.md

Local setup

1. Create a virtual environment

git clone <repo-url>
cd G
python -m venv .venv

Windows (PowerShell)

.venv\Scripts\Activate.ps1

macOS / Linux

source .venv/bin/activate

2. Install dependencies

pip install -r requirements.txt

3. Configure environment variables

Copy .env.example to .env and add your Groq API key:

GROQ_API_KEY=your-groq-api-key-here

4. Run the app

streamlit run src/app/streamlit_app.py

5. Run tests

pytest tests/ --tb=short -q

Data and methodology notes

  • Source data: data/option_a_ticket_data.csv
  • Comparable weeks: 7–10
  • Week 11 is partial and excluded from week-over-week interpretation
  • Raw data is not modified in place
  • Opportunity sizing uses explicit scale assumptions from the project config and analysis

Notes for reviewers

If reviewing quickly, the best path is:

  1. Open the app in src/app/streamlit_app.py
  2. Review the EDA in notebooks/eda_analysis.ipynb
  3. Inspect the agent flow in src/agent/graph.py
  4. Check the tests in tests

Limitations

  • The dataset is synthetic and intentionally noisy
  • Annualized impact depends on stated scale assumptions
  • LLM-backed stages depend on API availability and rate limits
  • The current version is a prototype, not a production deployment

About

No description, website, or topics provided.

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages