ThendralMagudapathi

Building enterprise Generative AI systems for SEC filing analysis using LLMs, RAG and multi-agent architectures.

• Built enterprise RAG pipelines for SEC filing intelligence.
• Fine-tuned LLMs using LoRA, QLoRA and PEFT.
• Developed Context Retrieval and Multi-Agent AI workflows.
• Integrated MCP, Ollama, LangChain and Vector Databases.
• Optimized production inference for scalable enterprise AI.

• Python
• FastAPI
• Large Language Models
• RAG
• Agentic AI
• LangChain
• FAISS
• SQLite
• Docker
• REST APIs
• Git

Engineered production Computer Vision, Edge AI and Generative AI systems across fitness analytics, industrial safety and intelligent monitoring platforms.

• Built AI-powered fitness tracking using pose estimation, activity recognition and video understanding.
• Developed person tracking, face re-identification, object detection and segmentation pipelines.
• Implemented spatio-temporal action recognition using SlowFast, LSTM and Transformer architectures.
• Designed YOLO-based PPE detection on NVIDIA Jetson Orin Nano with RGB + Thermal camera fusion.
• Achieved 90% accuracy for thermal-CCTV intrusion detection in industrial environments.
• Developed thermal body temperature estimation and environmental anomaly detection models.
• Built a personalized RAG assistant integrating LLMs, FAISS, LangChain and Vector Databases.
• Optimized IoT sensor analytics using FFT, FIR, IIR and Savitzky-Golay signal processing.
• Led and mentored a team of 5 AI interns while collaborating with product and design teams.

• Python
• PyTorch
• OpenCV
• YOLO
• CUDA
• TensorRT
• NVIDIA Jetson Orin Nano
• SlowFast
• LSTM
• Vision Transformers
• CNN
• Face Re-identification
• Object Detection
• Image Segmentation
• Pose Estimation
• Video Understanding
• Activity Recognition
• Thermal Imaging
• RAG
• LLMs
• LangChain
• FAISS
• Vector Databases
• Docker
• FastAPI
• Linux

Conducted applied AI research in Generative AI and Physics-Informed Neural Networks.

• Developed YOLO-based PPE detection on NVIDIA Jetson Orin Nano.
• Built RGB + Thermal fusion models for industrial safety.
• Achieved 90% accuracy for intrusion zone detection.
• Developed thermal body temperature estimation using segmentation.
• Designed IoT temperature anomaly detection with automated alerts.
• Led and mentored a team of 5 AI interns.
• Applied FFT, FIR, IIR and Savitzky-Golay filters for sensor optimization.

• Python
• PyTorch
• Deep Learning
• Computer Vision
• GANs
• Conditional GANs
• Physics-Informed Neural Networks
• DeepXDE
• Neural Networks
• Data Analysis

Conducted applied AI research in Generative AI and Physics-Informed Neural Networks.

• Developed GAN and CGAN models for 1D & 2D ultrasonic sensor data.
• Achieved 90% (1D) and 70% (2D) model accuracy.
• Implemented Physics-Informed Neural Networks for heat transfer simulation.
• Designed custom Generator, Discriminator and U-Net architectures.
• Mitigated GAN mode collapse using training stabilization techniques.

• Python
• PyTorch
• Deep Learning
• Computer Vision
• GANs
• Conditional GANs
• Physics-Informed Neural Networks
• DeepXDE
• Neural Networks
• Data Analysis

Been a part of building CAD software at IIT Madras Research Park using Python and Open Cascade Technology.

• Built 10 reusable CAD modules using Python and OOP.
• Developed backend components using Open Cascade Technology (OCCT).
• Implemented UI integration for CAD product workflows.
• Improved code modularity through reusable software architecture.
• Delivered production-ready engineering modules from scratch.

• Python
• Open Cascade Technology
• PyQt
• Qt Designer
• SQLite
• Object-Oriented Programming
• Git