Research Profile
Works on structural alignment, computation, and representation learning to understand large-scale systems with weak supervision, particularly in settings with noisy labels, ambiguous targets, and distribution shift. Experience includes training and evaluating classifiers on image and social media engagement datasets, applying regression to ambiguous targets, and building pipelines for scalable model development and validation. Interested in probabilistic modeling, active inference, and hierarchical representation learning, with a focus on bridging theory (uncertainty, structure, invariance) and applied ML systems.
Research Interests
Representation learning under weak and noisy supervision
Studying how models learn meaningful representations from imperfect labels and high-dimensional inputs, with an emphasis on extracting stable signals from noisy visual and user-generated data.
Learning under distribution shift and non-stationarity
Understanding how model performance evolves when data distributions change over time, and developing approaches that maintain reliability in dynamic environments.
Open-World Alignment & Non-Ergodic Dynamics
Formalizing structural boundaries where autonomous agents experience critical drift due to unmodeled environmental variables and path-dependent, non-linear feedback loops.
Continuous Belief-State Architectures
Moving past episodic calculation layers toward persistent tracking paradigms optimized for continuous, real-time prediction error minimization fields.
Technical Skills
Programming: Python, TypeScript, React
ML Methods: Supervised Learning, Regression, Classification, Feature Extraction
Systems Architecture: Structural Alignment & Computation, ML pipelines, workflow automation, rule-based systems, model evaluation & validation
Experience
Associate, Trust & Safety — Meta
Hyderabad, India- Worked on training and evaluation of machine learning classifiers for detecting harmful and adversarial content in large-scale social systems.
- Integrated image-based detection pipelines, dealing with high-dimensional, sensitive visual data under noisy labeling conditions and weak supervision.
- Addressed ambiguity by designing verification processes to improve the statistical consistency of model-assisted decisions.
- Participated in infrastructure efforts to mitigate misinformation and coordinated adversarial behavior, requiring real-time adaptation to highly non-stationary data distributions.
- Operated at the intersection of human-in-the-loop validation matrices and probabilistic ML models, refining feedback loops for improved performance.
Delivery Associate Engineer — Mphasis
Bangalore, India- Designed rule-based decision systems (Sigma rules) that formalized implicit system knowledge into highly structured automated workflows.
- Built early-stage automation pipelines, bridging deterministic logic with scalable system behavior to stabilize open-world inputs.
- Improved system efficiency by reducing operational uncertainty in escalation handling through structured prioritization matrices.
Director & Systems Operator — Artisanal Furniture Boutique
Chennai, India- Managed end-to-end supply chain logistics, inventory architectures, and specialized structural configurations for high-value physical assets.
- Coordinated multi-tier client deployments and precise physical product installations under variable, real-world environmental constraints.
Selected Projects & Manuscripts
Representation Learning for Image Classification
Developed supervised learning models to learn representations from image data. Implemented preprocessing and feature extraction pipelines for high-dimensional visual inputs and evaluated model performance under noisy or weak labeling conditions with a focus on robust signal extraction.
Scalable ML Pipeline Design
Built end-to-end pipelines for data ingestion, transformation, model training, and evaluation. Structured workflows to support iterative experimentation, reproducibility, and modular integration of models into larger decision systems.
Continuous Cognitive Architecture Portfolio (Manuscripts)
• "The Gap Between Computation and Cognition" — Structuring unpausing state-space systems.
• "The Postulate of Continuity" — Modeling continuous world-model persistence metrics.
• "World Models: The Next Representation" — Shifting primitives to simulation fields.
Education
Post Graduate Diploma in Management — Xavier Institute of Management & Entrepreneurship, Bangalore
B.Tech in Information Technology — Meenakshi Sundararajan Engineering College, Chennai