• 🎓 MSc Data Science — CHRIST (Deemed to be University), Bengaluru (2024–2026)
• 💼 AI/ML Engineer Intern at TechZia — building autonomous vehicle cybersecurity systems with CNN + LSTM + MARL
• 📄 IEEE-Published Researcher — ICVADV-2026, IEEE Computational Intelligence Society
• 🏛️ Government of India Design Patent Co-Holder — Mobile Cooler, Class 23-04, No.: 443142-001
• 🤖 Proficient in end-to-end ML/DL pipelines, NLP, LLMs, RAG, AI Agents, and Generative AI
• 📊 Experienced in EDA, statistical modelling, KPI reporting, and BI dashboards
• ⚡ Fun fact: I benchmark generative models for astronomical datasets and build RL agents for cybersecurity — before breakfast

Programming Languages: Python | R | SQL | Java | JavaScript

ML & Deep Learning: Scikit-learn | TensorFlow | Keras | XGBoost | CNN | LSTM | GRU | Transformers | GANs | VAE | Diffusion Models | GNN | LLMs | LangChain | RAG | AI Agents | Prompt Engineering

Data & Libraries: Pandas | NumPy | Matplotlib | Seaborn | Plotly

Visualisation & BI: Power BI | Tableau | KPI Reporting | Dashboard Development

Databases, Cloud & Tools: PostgreSQL | MySQL | AWS | Azure ML | Flask | FastAPI | Streamlit | Docker | Git | GitHub

Autonomous Vehicle Cybersecurity Agent

• Architected a CNN + LSTM real-time anomaly detection pipeline on AV telemetry (IMU, CAN bus, GPS) within ROS 2 Humble, enabling continuous inference across 3 concurrent sensor streams
• Engineered a DQN-based Multi-Agent RL (MARL) framework with 3 specialised agents (sensor security, vehicle behaviour, network cybersecurity) orchestrated by a hierarchical global decision agent
• Preprocessed the comma.ai AV dataset via vectorised NumPy pipelines; researched CNN-Transformer hybrid and GNN-based multi-sensor fusion architectures for vehicle network intrusion detection

Synthetic Data Augmentation for Robust Solar Flare Classification: Comparative Analysis of Conditional DCGAN, VAE and Diffusion Models

• Benchmarked 3 generative architectures (Conditional DCGAN, VAE, Diffusion Models) for synthetic image generation
• Improved classifier robustness on imbalanced astronomical datasets

Mobile Cooler — Class 23-04 — Patent No.: 443142-001

• Applied CNN transfer learning on 10,000+ images across 90 species; boosted accuracy 35% → 91.76% via hyperparameter tuning (learning rate, batch size, dropout)
• Deployed production REST API (Flask) handling 100+ req/sec at <200ms inference latency

• Built DenseNet121 + GRU + Attention spatio-temporal model on 8,000+ satellite sequences; achieved 75% accuracy, 58% M-class and 36% X-class rare event recall
• Deployed client-side via TensorFlow.js in React — zero backend infrastructure required

• Engineered a Profit Margin column and Risk Score metric to flag loss-making transactions from retail transaction data
• Built an interactive Power BI dashboard with KPI cards (Total Sales, Profit, Orders, Avg Margin), category-wise sales, regional profit, and dynamic slicers
• Ran SQL aggregations to surface actionable insights for pricing strategy and inventory optimisation

📧 Email: jestinthomas1507@gmail.com 🔗 LinkedIn: linkedin.com/in/jestin-thomas-90a109317 🐙 GitHub: github.com/JestinThomas 📍 Location: Bengaluru, India