Karan Mittal

  • Data visualisation expert
  • Co-founder and CTO, Dextar
  • Organiser, ML Indore and GDG Cloud Indore

I’ve built data apps in Plotly Dash since 2019, including a year as a software engineer at Plotly. After that I built analytics for semiconductor manufacturing at Lam Research and for oncology teams in life sciences.

Discuss your projectWriting and talks
Karan Mittal

What I build

Most of my work starts with data that’s hard to read and ends with something a team opens every week.

  1. A weekly metric against its targetIllustrative line chart of orders shipped on time over twelve weeks. The line starts at 84%, stays under the 90% target until week 10 and ends at 94% in week 12.On-time orders94%Target 90%Week 1Week 12

    Dashboards your team can run

    Plotly Dash and Streamlit apps on your own data, built so your team can change them after I leave.

  2. Evaluation score across model versionsIllustrative line chart of an evaluation score for model versions 1 to 7. The score rises from 0.79 to 0.89 at version 5, drops to 0.76 at version 6, a regression, and recovers to 0.87 at version 7.Eval score0.800.90Regressionv1v2v3v4v5v6v7

    Charts that explain a model

    Evaluation results and errors drawn so people outside the ML team can see where a model fails.

  3. A path through connected recordsIllustrative network of ten records. One path is highlighted: from a customer to an order, from the order to a lot and from the lot to a supplier. The other six records are connected but not on the path.Connected recordsCustomerOrderLotSupplier

    Views of connected data

    Knowledge graphs and networks laid out so you can follow one record to the next.

The charts use illustrative data.

Writing and talks

All writing and talks

What needs to work better?

Describe the data, who relies on it and what they need to see. A few sentences are enough to start.

Start a conversation