Wanted to share a project I've been working on for the past two weeks.
The Problem
Every Monday, our ops team receives 5-6 CSV exports from different systems. Someone manually combines them, cleans the data, runs analysis, and produces a weekly report. Takes ~4 hours.
The Solution
I built an automated pipeline using Nuveki:
Step 1: Ingestion
Upload all CSVs to a conversation with Data Processor. It parses each file, identifies column types, and flags data quality issues (missing values, format inconsistencies).
Step 2: Cleaning & Merging
Same conversation — ask Data Processor to:
- Standardize date formats across all files
- Fill missing values using the rules I specified
- Merge on the common
account_id column
- Output a clean, consolidated CSV
Step 3: Analysis
Switch to a Team conversation with Data Processor + Content Summarizer:
- Calculate week-over-week changes for key metrics
- Identify outliers (>2 standard deviations)
- Generate trend descriptions in plain English
Step 4: Report
Ask the team to format everything into a structured report with:
- Executive summary (3 bullet points)
- Metrics table with RAG status indicators
- Detailed findings with charts described in markdown
- Action items
Results
- Time: 4 hours → 25 minutes (including my review)
- Consistency: No more "oh I forgot to clean that column" mistakes
- Flexibility: Easy to add new data sources — just upload and adjust the prompt
The key insight was breaking it into clear steps rather than trying to do everything in one prompt. The agents are remarkably good at each step individually.
Happy to share my prompt templates if anyone wants to replicate this.