Data Scientist & AI Researcher

BHARATH
KESAV R

Int. M.Sc. Data Science student at Amrita Vishwa Vidyapeetham building production-grade AI systems in Sports Analytics, Explainable AI, RAG pipelines, and Graph Neural Networks — 3 live deployed apps, 9 public repos.

Scroll
0 GitHub Repos
0 Live Deployments
0 Domains Covered
0 Graduating Year
About Me
WHO I AM

I'm an Integrated M.Sc. Data Science student at Amrita Vishwa Vidyapeetham, Coimbatore (2022–2027). My work sits at the intersection of deep learning, sports analytics, and explainable AI.

I build systems that don't just predict — they explain. From graph-based cricket outcome prediction to RAG-powered UCL match summaries, every project is designed around novelty, rigour, and real-world utility — with 3 live deployed applications to prove it.

My passion for cricket and football drives my research domains, where I apply IoT sensor fusion, Graph Attention Networks, Transformer architectures, LLMs, and SHAP explainability to extract actionable insights from raw data.

Backed by 9 public GitHub repositories, I combine publication-quality research with production-ready deployment using Streamlit and HuggingFace Spaces.

Degree Int. M.Sc. Data Science
Institute Amrita Vishwa Vidyapeetham, Coimbatore
Reg. No. CB.PS.I5DAS22110
Focus Graph DL · NLP · IoT · Sports AI · XAI
Location Coimbatore, Tamil Nadu, India
Email bharathkesav1275@gmail.com
Work
PROJECTS
01 Graph DL · Cricket
CricketGraph-DL

Spatio-temporal graph learning for IPL T20 match outcome prediction. GAT player-interaction graph + BiLSTM + cross-attention Transformer. Ball-by-ball win probability, run forecasting & Player Impact Score.

PyTorchPyGSHAPOptunaIoT
View on GitHub →
02 NLP · Transformers
Sarcasm Detection

Multi-task RoBERTa with novel Sentiment Incongruence Auto-Labeler. Dataset-agnostic labeling from semantic mismatch between surface sentiment and underlying emotion. F1-Macro 0.977, AUC 0.997.

RoBERTaFocal LossLIMEHuggingFace
View on GitHub →
03 IoT · Football
FootballRole-DL

Classifies football players into Attacker, Midfielder, Defender from PAMAP2 IoT wearable data. Activity-to-role mapping. M3 TCN+Transformer achieves 99.24% accuracy. LOSO 98.89%±0.42%.

TCNTransformerSHAPPAMAP2
View on GitHub →
04 IoT · ML
Football Fatigue Prediction

Three-class fatigue prediction from PAMAP2 wearable IoT data. Novel Karvonen heart rate labeling (leakage-free). Personalized Random Forest: 97.87% accuracy, LOSO 97.96%. Coach substitution-alert dashboard.

scikit-learnSMOTELOSOTensorFlow
View on GitHub →
05 RAG · LLM
Explainable Match Summaries

Factually grounded UEFA Champions League match summaries using RAG + LLaMA 3.1 via Groq API + SHAP. Self-curated UCL-2025 dataset (189 matches, 142 cols). Sentence-BERT + FAISS retrieval: cosine similarity 0.903 vs 0.373 baseline.

LLaMA 3.1Groq APIFAISSSHAPSentence-BERT
View on GitHub → 🚀 Live Demo →
06 ML · Sports Analytics
Football Match Prediction

Multi-league framework across 5 European leagues. ELO ratings + rolling form + betting market probabilities + H2H stats. 33 features, 3-layer anti-leakage architecture. Gradient Boosting: 0.55 acc, 0.91 xPts MAE. 259 teams available.

Gradient BoostingSHAPELOTimeSeriesSplit
View on GitHub → 🚀 Live Demo →
07 Reinforcement Learning
Goalkeeper RL

AI goalkeeper trained via PPO in a 2D Pygame football simulation. Rule-based threat assessment + shot prediction + state machine (IDLE→TRACKING→READY→DIVING→RECOVERING). Improves with every game.

PPOPygamePyTorchActor-Critic
View on GitHub →
08 Agentic AI
Smart Resource Finder

Agentic AI study assistant for college students. Observe→Think→Act loop using Groq API + LLaMA 3.3-70B. Finds curated resources and generates personalized exam study plans with live progress tracking.

LLaMA 3.3GroqStreamlitTool Calling
View on GitHub → 🚀 Live Demo →
09 NLP · Web App
Smart Study Helper

AI-powered web platform for students with ADHD, dyslexia & learning disabilities. T5 PDF summarization, spaCy quiz generation, text-to-speech, multilingual translation (7 Indian languages). Built with Django + PostgreSQL.

DjangoT5spaCyPostgreSQL
View on GitHub →
Expertise
SKILLS
Deep Learning & AI
PyTorch Graph Neural Networks Transformers LSTM / BiLSTM TCN GAT Reinforcement Learning PPO Multi-task Learning Focal Loss
NLP & LLMs
HuggingFace RoBERTa T5 LLaMA 3.1 RAG FAISS Sentence-BERT Groq API spaCy LIME Prompt Engineering
ML & Data Science
scikit-learn SHAP Gradient Boosting Random Forest SMOTE LOSO CV TimeSeriesSplit Optuna ELO Ratings Feature Engineering
Tools & Technologies
Python PyTorch Geometric TensorFlow Django PostgreSQL Kaggle GPU LaTeX Git / GitHub Streamlit Pygame
Get In Touch
LET'S
CONNECT

Open to research collaborations, internships, and exciting projects in AI, sports analytics, and deep learning. Feel free to reach out!

GitHub Profile →