DEV Community: Rajiv Sambasivan

New Features with TSEDA - get the most out of your time series data.

Rajiv Sambasivan — Thu, 07 May 2026 07:37:12 +0000

Update: Automating Time Series Exploration with tseda 📈

A while back, I shared tseda, a tool designed to help you make sense of high-frequency business metrics (like hourly conversion rates or service windows).
Since then, I’ve been working on making the transition from "collecting data" to "understanding data" even faster.

What’s New in tseda?

Automatic Window Management: You no longer have to guess your window sizes. The tool now handles automatic window size assignment and refinement, finding the "signal" in your data without the trial and error.
Notebook Parity: You can now move seamlessly between the tool and Jupyter notebooks. Keep your flow state intact while switching from visual exploration to deep-dive coding.

Why use it?

If you have data at an hourly or greater cadence, you’re likely looking for two things: Forecasting and Anomaly Detection. tseda is built to help you build better apps by actually understanding the underlying patterns of those metrics.

Get Started (or Catch Up):

New README & Docs: github.com/rajivsam/tseda
User Guide: Step-by-step instructions
Video Overview: AI-generated summary

I’m looking for feedback from anyone monitoring metrics at a high cadence. How are you currently handling window refinements? Let’s discuss in the comments!

Would you like me to tailor the technical highlights to focus more on the Markov analysis or the specific Python libraries you used?

KMDS with New Features

Rajiv Sambasivan — Mon, 27 Apr 2026 06:45:53 +0000

A little while ago I developed a python package meant for small data science teams to communicate the rationale and motivation for decisions in developing and modeling data science projects. The package was called KMDS. This package has an upgrade now. You can input your observations in natural language and the package will take care of tagging it appropriately based on a data science project ontology. Conversely, the natural language search is also available, you can query this tool in natural language.
The updated repository with examples is available here:
https://github.com/rajivsam/kmds
Thank you
Rajiv

Explore Your Time Series Data

Rajiv Sambasivan — Mon, 27 Apr 2026 06:40:44 +0000

Do you have business data that you collect at an hourly or greater cadence - for example, site conversion rate per day, average service time for the 10 am - 11 am window etc.? Do you want to understand this data so that you can build better apps based on your understanding - for example, forecast the metric you are monitoring for the next business period, understand if a particular value is anomalous.
If this is of interest to you, check out tseda, a tool to explore and understand your time series data
https://github.com/rajivsam/tseda
There is a video (AI summary) here:
https://www.youtube.com/watch?v=baoJrIpSTE8
There is a user guide here:
https://github.com/rajivsam/tseda/blob/main/docs/user_guide.md
I am happy to answer questions and discuss how you can use this if you are collecting a metric at an hourly cadence or higher.
I am making a version of this high frequency sensor data.
The technical motivation is available here: https://rajivsam.github.io/r2ds-blog/posts/markov_analysis_coffee_prices/

Descriptive Analytics

Rajiv Sambasivan — Wed, 04 Feb 2026 03:12:11 +0000

Descriptive Analytics is a repository that is a collection of recipes for descriptive analysis of enterprise data. It is work in progress. Integration with generative AI tools to combine conventional ML analysis techniques with generative AI tools is ongoing.
See https://www.youtube.com/watch?v=MwXKC_oloH8 for an overview, see
https://github.com/rajivsam/descriptive_analytics for the repository

KMDS, a package for knowledge managment in data science

Rajiv Sambasivan — Thu, 18 Apr 2024 08:17:31 +0000

KMDS is a tool that solves a problem that most folks doing data analysis are frustrated with. You run into a design question, you know you've dealt with this in the past, you just can't recreate the context, question, research and the rationale for picking a solution when you run into the problem next. Here is a new release of the tool with examples of how you use it for both machine learning and analytics workflows. See https://github.com/rajivsam/kmds. Here is a short video description: https://www.youtube.com/watch?v=n7gE6bfLWtI