Skip to content
View rajivsam's full-sized avatar

Block or report rajivsam

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
rajivsam/README.md

Rajiv Sambasivan

Statistical learning for data-rich companies that want ML they can understand, own, and afford

Most ML problems I'm called in on aren't model problems. They're data problems (hidden heterogeneity, shifting regimes, the wrong window) or ownership problems (a system nobody on the team can explain or maintain). I fix both, with PhD-level statistical training and 25+ years of production software engineering behind the work, and without the big-vendor premium.

Best fit if you:

  • have an existing ML system that is hard to interpret, expensive to run, or fragile to maintain, or
  • are starting a new ML initiative and want it built to be explainable and owned by your team from day one.

How I work

  1. Diagnose before modeling. Structured audits of your data: trends, seasonality, structural breaks, and whether one model is even the right answer.
  2. Prefer interpretable methods. Classification, regression, survival analysis, panel data, and optimization, with the simplest approach that meets the need.
  3. Hand over what you own. Clean Python packages, documented and reproducible, no vendor lock-in, no black boxes your team can't maintain.

What this looks like in practice

Your situation What I do Typical outcome
Model performance has drifted and nobody knows why Regime/structural-break diagnostics see tseda and structural decomposition tool for examples
Operational counts (orders, claims, tickets) swing unpredictably Count Time Series Analysis see count window selector for examples
One model underperforms across customer segments Heterogeneity testing and structure identification [e.g., segment-aware design], see structural decomposition tool
Cloud/vendor ML costs are climbing Right-size to simpler, in-house-maintainable methods see the kmds examples for how a solution would look

Open-source toolkit (what I use on engagements)

Tool Solves
tseda Decompose a time series into trend, seasonality, and noise before you model it
count_ts_window_selector Detect real rate changes in count data
structure_id Diagnose heterogeneity in cross-sectional data
KMDS Keep data science work documented and reusable across a team (case studies)

For an example of typcial project development summary, see the sba model development summary

Writing

Background

PhD in Machine Learning (Google Scholar) · 25+ years across financial services, retail, telecom, transportation, and government.

Let's talk

Start with a short conversation about your current ML setup or planned project: r2-ds.com LinkedIn · Substack

Popular repositories Loading

  1. tseda tseda Public

    tool for exploring regularly sampled time series data

    Python 15

  2. KMDS KMDS Public

    Source for KMDS

    Jupyter Notebook 6

  3. LSH-Minhash- LSH-Minhash- Public

    LSH implementation using Minhash on congressional voting record dataset

    Python 2

  4. itsm_retail_examples_r2ds itsm_retail_examples_r2ds Public

    notebook and data folders for r2ds itsm and retail

    Jupyter Notebook 2

  5. kmds_recipes kmds_recipes Public

    Repository of kmds recipes for common kmds tasks

    Jupyter Notebook 2

  6. ICA ICA Public

    Independent Component Analysis of the glass dataset using sklearn's implementation of Fast ICA

    Python 1 2