Tijo Thomas · Data Scientist — India

Machine learning that makes it out of the notebook.

I build ML pipelines, LLM agents, and the infrastructure that keeps them alive in production — from raw data to deployed model, reproducibly.

See the work GitHub ↗

fig. 0 — validation loss. converging since 2019.

fig. 1

Selected work

Four builds, each taken all the way — from raw data to something you can deploy, version, and reproduce.

Agentic LangGraph Chatbot

A chat agent that plans multi-step answers and calls live tools instead of one-shot prompting — built on LangGraph's stateful graphs.

Python · LangGraph · LLM agents

US Visa Approval Prediction

An end-to-end ML system: ingestion, validation, training, and deployment as one reproducible pipeline. The full MLOps loop, not just a model in a notebook.

Python · scikit-learn · MLOps

Pneumonia Detection

A convolutional network trained on thousands of labeled chest X-rays, shipped as a web app so a prediction is one upload away — deep learning with a front door.

Deep learning · Computer vision · Web

ML Pipeline with DVC + MLflow

Experiment tracking and data versioning done properly: every run reproducible, every model traceable back to the exact data and code that produced it.

DVC · MLflow · Reproducibility

fig. 2

Skills, plotted

Read this the way you'd read a legend — each line is a series I work in daily.

Machine & deep learning scikit-learn · TensorFlow · PyTorch · classification · regression · time series
Generative AI & agents LangChain · LangGraph · smolagents · RAG · prompt & tool design
MLOps Docker · MLflow · DVC · CI/CD · model deployment & monitoring
Cloud AWS (S3 · EC2 · ECR) · Azure
Data Python · Pandas · NumPy · SQL · EDA · visualization

fig. 3

Trajectory

Time on the x-axis, like any honest chart.

2019
First commit. Web fundamentals — HTML, CSS, JavaScript, and the habit of shipping.
2021
Web development internship — built and deployed an NGO donation site with payments.
2023
Deep into ML & DL — classification, regression, NLP, computer vision, time series.
2024
MLOps era — Docker, DVC, MLflow, and pipelines that survive without their author.
now
Data scientist at Lechler India — building LLM agents and RAG systems for engineering workflows.

fig. 4

References

Every good figure cites its sources. Here are mine.

[1] T. Thomas, GitHub — github.com/Tijo2000
[2] T. Thomas, LinkedIn — linkedin.com/in/tijo-thomas
[3] Correspondence — tijo.thomas@lechlerindia.com

Get in touch