Sai Teja Lodagala
Graduate engineer specializing in Model Predictive Path Integral (MPPI) Control, Nonlinear Dynamic Systems, Humanoid Locomotion, Physics-Informed Deep Learning, and Psychoacoustic Audio Optimization.
Sai Teja Lodagala
IIT Bombay
Systems & Control Engineering
Education
Postgraduate & undergraduate engineering training.
M.Tech in Systems & Control Engineering
Indian Institute of Technology Bombay (IIT Bombay)
- Specialization in Systems & Control Engineering with research in Intelligent Audio Quality Optimization under Prof. Raj Anguluri.
- Focus on Model Predictive Control, Optimal Control, Machine Learning, Digital Signal Processing, and Non-linear Dynamics.
B.Tech in Electronics & Communication Engineering (ECE)
Anil Neerukonda Institute of Technology and Sciences (ANITS), Andhra University
- Core training in Signals & Systems, Control Systems, Digital Communication, and Microprocessors.
- Capstone Project: Farm Monitoring & IoT Crop Health Anomaly Detection System using CNNs.
Projects & Research
M.Tech research at IIT Bombay, open-source control suites, robotics platforms, and deep learning architectures.
Swar Studio (Intelligent Audio Quality Optimization Platform)
Real-Time Neural Denoising, 4-Stem Separation & Hearing-Inspired Psychoacoustics
- Engineered a real-time neural audio denoising and 4-stem source separation platform utilizing Complex Spectrogram UNet architectures, achieving processing latency under 20 milliseconds.
- Implemented dynamic audio mastering protocols integrating hearing-inspired psychoacoustic ISO 226 contours for artifact-free and perfectly normalized broadcast-quality loudness.
- Architected high-throughput cloud-native REST and WebSocket API endpoints using FastAPI and Kubernetes, facilitating highly scalable live-stream audio processing.
Autonomous Vehicle ML-MPC & MPPI Framework
Physics-Informed Deep Residual Dynamics & Batch-Vectorized MPPI Control with Pacejka Tire Physics
- Engineered a 6-DOF planar dynamic bicycle model integrating Pacejka Magic Formula tire curves with Runge-Kutta 4th Order (RK4) numerical integration.
- Implemented batch-vectorized MPPI controller in PyTorch evaluating 5,000 candidate trajectories per control cycle with non-convex obstacle repulsion fields.
- Designed Deep Residual MLPs with LayerNorm and GELU to dynamically compensate for unmodeled aerodynamic drag and surface friction variations.
Bipedal Locomotion & Humanoid Control Suite
5-Link Planar Underactuated Biped (HZD) & 3D LIPM with ZMP Preview Control in Simscape
- Derived complete 5-link Lagrangian equations of motion with discrete heel-strike impact reset maps and Poincaré limit-cycle stability analysis.
- Implemented input-output feedback linearization enforcing virtual constraints (Hybrid Zero Dynamics) for stable, periodic underactuated planar walking.
- Developed 3D Linear Inverted Pendulum Model (LIPM) with discrete Riccati ZMP preview control and 12-DOF inverse kinematics in Simscape Multibody.
Inverted Pendulum Cart-Pole Control Suite
Nonlinear Lagrangian Modeling, Dual-Loop PID, LQR Optimal Control & LQG Kalman Filter
- Derived first-principles nonlinear equations of motion using Euler-Lagrange mechanics and performed Taylor series Jacobian state-space linearization.
- Designed and tuned Dual-Loop Cascaded PID and Linear Quadratic Regulator (LQR) with feedforward tracking for cart regulation and pole stabilization.
- Implemented Linear Quadratic Gaussian (LQG) controller integrating a Kalman filter for optimal state estimation under measurement noise.
V2V Communication & Connected Vehicle Platooning
Cooperative Adaptive Cruise Control (CACC), String Stability & Distributed MPC
- Engineered Cooperative Adaptive Cruise Control (CACC) proving L2 string stability across all frequencies, enabling safe platooning at ultra-tight headways (h = 0.3s).
- Implemented Distributed MPC (D-MPC) with state extrapolation algorithms compensating for wireless network transmission latency and packet loss.
- Designed a Chi-Square residual anomaly detector to identify and isolate False Data Injection (FDI) cyber-attacks on vehicle telemetry.
BLDC Motor Speed Control with PWM & Hall Commutation
6-Step Hall Commutation, 20 kHz PWM Chopping & Closed-Loop Anti-Windup PI Control
- Modeled 3-phase BLDC stator electrical dynamics and trapezoidal back-EMF with 120° electronic commutation and Hall sensor sector decoding.
- Implemented high-side 20 kHz PWM chopping for 3-phase Voltage Source Inverter (VSI) power stage.
- Designed a closed-loop PI speed controller equipped with dynamic anti-windup clamping to eliminate integrator saturation under step load disturbances.
Omni Code (AI Code Collaboration Platform)
Autonomous Engineering Workspace: Distributed Code Management × Living Specs × Cursor AI
- Engineered an autonomous engineering workspace integrating distributed code management and interactive PR reviews with real-time generative LLM intelligence to accelerate developer workflows.
- Architected a semantic natural language search engine and context-aware auto-documentation generator, ensuring repository architectures and living specifications remain consistently synchronized.
- Developed a one-click visual diff application mechanism leveraging continuous underlying AST diagnostics to instantly deploy LLM-suggested architectural modifications.
Fraud Sentinel AI (Dual-Engine Fraud Detection System for Banking)
Cross-Channel Threat Fusion: NLP Smishing Detection × Behavioral Transaction Anomaly Scoring
- Designed a robust cross-channel threat fusion engine seamlessly correlating NLP-driven SMS smishing and phishing detection with behavioral bank transaction anomalies.
- Deployed a dual machine learning pipeline combining Isolation Forest and Random Forest algorithms, executing automated account quarantines based on real-time risk thresholds.
- Constructed an interactive Streamlit SOC dashboard continuously streaming live telemetry and behavioral baselines to visually simulate kill-chain events.
Adversarial Examples in Neural Networks (Reproducing FGSM)
Empirical Reproduction of Goodfellow (ICLR 2015) FGSM, 8 Attack Vectors & Defense Zoo
- Developed a production-grade research platform dedicated to the empirical study of neural networks, featuring eight sophisticated adversarial attack vectors and four model defenses.
- Built an interactive FastAPI web studio equipped with a dynamic real-time drawing canvas, detailed perturbation heatmaps, and responsive 2D loss landscape visualizers.
- Achieved an 81.25% robustness gain on the MNIST image dataset by systematically implementing adversarial training and randomized defensive Gaussian smoothing filters.
Farm Monitoring System (IoT & CNN Crop Health Analytics)
Agricultural Anomaly Detection with Embedded Sensors & CNN Leaf Disease Classification
- Developed an advanced agricultural anomaly detection system utilizing embedded IoT sensors to continuously track environmental shifts in soil moisture and pH with real-time alerts.
- Classified complex crop health conditions by deploying Convolutional Neural Networks on visual leaf data, subsequently generating precise chemical fertilizer prescriptions.
Twitter Sentiment Analysis using NLP
End-to-End NLP Pipeline with TF-IDF N-Grams, Calibrated Classifiers & Streamlit Dashboard
- Engineered an end-to-end NLP pipeline with regex text normalization, automatic hashtag segmentation, and emoji translation to heavily optimize social media text for sentiment classification.
- Trained mathematically calibrated, L2-regularized Logistic Regression and Multinomial Naive Bayes models on sparse TF-IDF vectors, consistently achieving over 94% test accuracy.
- Deployed an interactive Streamlit web dashboard facilitating real-time inference, token-level sentiment highlights, and batch CSV processing with Plotly visualizations.
Adversarial Minimax Game Engine (Tic-Tac-Toe)
Unbeatable Adversarial Search with Alpha-Beta Pruning, Pygame GUI & Console Engine
- Developed an unbeatable artificial intelligence engine utilizing recursive Minimax adversarial search with Alpha-Beta pruning to guarantee instantaneous optimal moves.
- Designed a versatile dual-interface architecture featuring a modern, responsive Pygame GUI alongside a lightweight, pure Python interactive terminal console.
- Built a comprehensively designed automated test suite specifically to validate complex board states and underlying game-theoretic optimality proofs.
Skills & Technologies
Modern control theory, machine learning frameworks, and infrastructure tooling.
Systems & Control Theory
Machine Learning & AI
Languages & Tools
Extra Curricular Activities
Competitive Achievements @ IIT Bombay
Pool Institute Hostel General Championship
IIT Bombay
Demonstrated strategic gameplay and consistent performance under pressure against top-tier collegiate participants.
Cricket Hostel 6 General Championship
IIT Bombay
Contributed to tactical success and athletic excellence in collegiate tournaments.
Interests
Passions Outside Engineering
Get in Touch
Feel free to reach out for research discussions, engineering collaborations, or opportunities in Model Predictive Control, Robotics, Machine Learning, Deep Learning, and Systems & Control.