Deepan Jayaraman

Industry

Four years building analytics that has to survive contact with messy data, incomplete information, and a decision that has to be made regardless. The methods come from the research; the constraints come from production.

Caterpillar India Engineering Solutions

Senior Associate Engineer — Data Science · Chennai · Jul 2023 – present

Generative AI and agent systems

Building Python-based generative AI tooling and agent workflows for analytics use cases across engineering solutions — the current focus of my work.

Engineering analyticsKnowledge retrieval
  • Python
  • LLMs
  • RAG
  • LangChain
  • Ollama

Predictive modelling and forecasting

Time-series predictive models feeding data-driven decision support for engineering and business planning.

Business planningEngineering planning
  • Python
  • Time series
  • LSTM
  • GAN

Machine-learning-enhanced digital twins

Digital twins for strategic planning, including battery thermal and ageing management — coupling physical models with learned components so the twin stays useful as conditions drift.

Battery thermal managementStrategic planning
  • Python
  • MATLAB

Decision-support systems and analytics dashboards

Interactive dashboards for motor design data, used across engineering teams to support design decisions rather than to report on them after the fact.

Motor designEngineering decision support
  • SQL
  • Power BI

Risk assessment and process optimisation

Applying probabilistic methods and advanced analytics to assess risk and optimise engineering processes — the doctoral work, in production.

ReliabilityProcess optimisation
  • Python
  • MATLAB

Onward Technologies

Data Scientist · Chennai · Jun 2021 – Jul 2023

Client-facing analytics delivery

Statistical analysis and data-driven decision support for industrial clients, working directly with them to shape and deliver their analytics roadmaps.

Industrial clientsDecision support
  • Python
  • SQL
  • Power BI

Data engineering and model selection

Building extract–transform–load pipelines, and identifying the modelling approach best suited to a given dataset and the question being asked of it.

Data pipelinesModel selection
  • Python
  • SQL
  • ETL pipelines

Production machine learning

Developing predictive models and algorithms, and implementing advanced machine-learning techniques in production settings.

Production MLPredictive modelling
  • Python
  • scikit-learn
  • XGBoost

Programme Director — Data Science for Industry

Designed and directed a two-day industry workshop, Chennai, December 2021.

TrainingIndustry outreach

Siemens

Research Associate · Sponsored project · 2017–2020

L-Moments for Statistical Analysis

Applied L-moment methods to industrial statistical analysis, strengthening the robustness of uncertainty quantification and risk assessment. Validated on real engineering problems including gas turbine disk design.

Where the research meets the work

The two halves are not separate tracks. The doctoral work asked how to make defensible decisions when data is scarce and contains extremes; industry supplied the cases where that is not a hypothetical — a batch of a few dozen measurements containing one genuine outlier, a failure mode with no historical precedent, a model whose recommendation someone has to sign off. My current research on prescriptive analytics and the model-to-decision handover came directly out of watching predictive models get handed to optimisers that inherited whatever the model happened to attribute.

The research programme →

Technical stack

Data lands and is shaped, models are fitted and explained, and the result is deployed and shown to someone who has to decide. Hover or tap an area to see the tools it uses.

Technical stack across six areas Six cards arranged around a central node labelled Technical stack: data engineering; modelling and machine learning; generative AI and LLMs; deployment and MLOps; visualisation and BI; and cloud. Each card names representative tools; the full list for every area follows below the figure. Technical stack Data engineering SQL · Spark · Snowflake Modelling & ML scikit-learn · XGBoost Generative AI & LLMs RAG · LangChain · Ollama Deployment & MLOps Git · CI/CD · Flask Visualisation & BI Power BI · Tableau Cloud Azure · AWS
The six areas the day-to-day work draws on. Every tool in each area is listed below.

Data engineering

  • SQL
  • ETL pipelines
  • Snowflake
  • Amazon Redshift
  • MySQL
  • Apache Spark

Modelling and machine learning

  • Python
  • R
  • MATLAB
  • scikit-learn
  • XGBoost
  • TensorFlow
  • Keras
  • LSTM
  • GAN

Generative AI and LLMs

  • Large language models
  • Retrieval-augmented generation
  • LangChain
  • Ollama

Deployment and MLOps

  • Git
  • GitHub
  • GitLab CI/CD
  • Flask
  • Model monitoring

Visualisation and BI

  • Power BI
  • Tableau
  • Matplotlib
  • Seaborn

Cloud

  • Microsoft Azure
  • Amazon Web Services

Professional development

AI Agents and GenAI for Enterprise Transformation

Centre for Outreach and Digital Education (CODE), IIT Madras, offered jointly with the IIT Madras FedEx SMART Center. December 2025 – February 2026.

Data Science and Artificial Intelligence — Leadership Essentials

Centre for Outreach and Digital Education (CODE), IIT Madras. January – March 2025.

Techie Award, Caterpillar India (2024) — outstanding technical contribution and excellence in engineering solutions.