Computational Neuroscience • Clinical AI • Machine Learning Systems

B.Sc. Computer Science — University of West London (RAK) Research Focus: Neural Decoding, Neuroimaging, Interpretable AI

My work lies at the intersection of machine learning and computational neuroscience, with an emphasis on building interpretable and clinically relevant AI systems. I am particularly interested in understanding how neural signals can be decoded, modelled, and translated into real-world applications such as brain–computer interfaces and neuroimaging diagnostics.

I develop end-to-end machine learning systems that integrate deep learning architectures, signal processing techniques, and explainability frameworks, with a strong focus on reproducibility and research-grade evaluation.

• Brain–Computer Interfaces (BCI) and Neural Decoding
• Computational Neuroimaging (MRI-based modelling)
• Affective Computing and Brain–Speech Interfaces
• Interpretable Machine Learning in Clinical Systems
• Neural Signal Processing and Time-Series Analysis

A Critical Evaluation of AI-Driven Neural Decoding Technologies
University of West London — Artificial Intelligence Module

Author: Pavithra Binu

This work presents a structured evaluation of five landmark brain–computer interface (BCI) studies published between 2006 and 2023, analysing the evolution of neural decoding systems from early intracortical interfaces to modern speech and motor restoration technologies.

Key Contributions

• Identifies a 4–5× increase in communication bandwidth (2019–2023) driven by a transition toward decoding internal motor representations.
• Traces the progression of decoding architectures from linear models and Kalman filters to deep recurrent networks and LLM-integrated systems.
• Synthesises findings across high-impact literature (Nature, The Lancet) to evaluate clinical readiness and system scalability.

Critical Gaps

• Long-term neural signal stability remains unproven (>5 years)
• Lack of participant generalisation across studies
• Minimal integration of sensory feedback (bidirectional systems)
• Limited standardisation of evaluation metrics
• Weak translation from laboratory prototypes to clinical deployment

97% Accuracy · 0.99 AUC-ROC

• Multi-class neuroimaging pipeline using EfficientNet, ResNet50, and Vision Transformers
• Grad-CAM-based interpretability for localisation of tumour regions
• Designed for reproducibility and extension in clinical research settings

99.1% Accuracy · AUC 0.999

• Ensemble-based disease risk prediction across oncology and cardiovascular domains
• Incorporates class imbalance correction (SMOTE) and feature engineering
• Evaluated using clinically relevant metrics (F1, AUC-ROC, precision/recall)

86% Accuracy · 0.85 F1-score

• CNN–BiLSTM architecture with attention for affective state classification
• Applications in affective neuroscience and brain–speech interface modelling

~400ms End-to-End Latency

• Streaming speech-to-intelligence pipeline (ASR → LLM → structured outputs)
• Demonstrates architectural parallels to real-time neural signal processing systems

Machine Learning & Deep Learning
PyTorch · TensorFlow · Scikit-learn

Architectures & Methods
CNNs · Vision Transformers · BiLSTM + Attention · Transfer Learning · Ensemble Methods

Scientific & Clinical Data
Medical Imaging (MRI) · Clinical Datasets · Audio Signal Processing

Systems & Infrastructure
FastAPI · Docker · PostgreSQL · WebSockets · Reproducible ML Pipelines

B.Sc. Computer Science
University of West London, RAK Branch Campus
2024 – 2027
GPA: 3.8 / 4.0

I am actively seeking opportunities to contribute to research in computational neuroscience and computational biology, particularly in areas involving neural data, brain–computer interfaces, and clinically deployable AI systems.

Email: [email protected]
LinkedIn: https://www.linkedin.com/in/pavithra-binu-55446a3a9/

Open to research collaborations and remote research assistant positions.