Silhouette facing a glowing ring of light

Karthikeyan Y

AI Engineer

I build models and AI systems that turn raw data into decisions — from classical ML to LLM-powered products.

Scroll

About

I turn messy data into models that hold up in production

Python and SQL as the foundation, strong DSA fundamentals, and hands-on work across machine learning, deep learning, NLP and generative AI — from classical models to LLM-powered systems.

I care about clean pipelines, measurable evaluation, and shipping models that teams can actually trust in production.

ML · DL
Core stack
NLP · RAG
Language AI
GenAI
Agents & LLMs

Neural network

Skills

Tools I use to ship reliable models

S/01

Python

Core language for data, modelling and tooling.

S/02

SQL

Query design, joins, window functions, warehousing.

S/03

DSA

Data structures and algorithms for clean, fast code.

S/04

Machine Learning

Regression, trees, boosting, evaluation.

S/05

Deep Learning

CNNs, RNNs, transfer learning with PyTorch.

S/06

NLP

Text pipelines, embeddings, transformers.

S/07

Generative AI

LLM apps, prompting, RAG, agents.

Dithered halftone texture

Always learning — currently deepening LLM evaluation and agentic workflows.

Selected work

Machine learning projects that solve real problems

01Python · scikit-learn · XGBoost

Customer Churn Predictor

End-to-end ML pipeline that scores subscription customers on churn risk, with feature engineering on usage logs and SHAP-based explanations for retention teams.

View code
02PyTorch · CNN · Transfer Learning

Medical Image Classifier

Deep learning model for chest X-ray classification using a fine-tuned ResNet backbone, augmentation and Grad-CAM heatmaps to surface where the model looks.

View code
03spaCy · Transformers · FastAPI

Resume Screening NLP Engine

NLP service that parses resumes, extracts entities and ranks candidates against a job description using sentence embeddings and cosine similarity.

View code
04LangChain · Vector DB · LLM

RAG Knowledge Assistant

Generative AI assistant that answers questions over private documents with chunked retrieval, citation of sources and streaming responses.

View code
05Pandas · Prophet · SQL

Sales Forecasting System

Time-series forecasting on multi-store retail data with holiday effects, backtesting and a SQL warehouse feeding scheduled retraining.

View code
06NLP · BERT · Streamlit

Sentiment Intelligence Dashboard

Fine-tuned BERT classifier over product reviews with aspect-level sentiment, topic clustering and a live dashboard for trend monitoring.

View code

Contact

Open to ML / AI roles and interesting collaborations