Projects & Experience

Mapped the Qwen2.5-3B residual stream to identify ~5–6D semantic submanifolds, establishing an empirical layer-selection criterion for safe steering.

Designed geometry-preserving steering operators that eliminate the 20× residual stream norm collapse inherent to standard additive CAA baselines.

Revealed RLHF safety routing occurs 18 layers before peak gradient saliency, proposing an entropy-delta proxy for accurate behavioral targeting.

Working to improve the 72% accuracy achieved on the Synthetic European dataset by refining KAN architecture.

Tracked and optimized model performance: train loss 2.84, test loss 4.57.

Managed regularization (±2.05) to improve model stability and generalization.

Machine Learning Intern, Glance

Designed an LLM-as-judge evaluation pipeline for catalog recommendations, testing prompting strategies (chain-of-thought, structured rubrics, multi-step verification) to improve judgment reliability and reduce false positives.

Developed a method to distill human annotator and LLM-judge signals into a cross-encoder, enabling recommendation scoring for zero-usage / cold-start users with no behavioral history.

Built and optimized catalog retrieval pipelines, improving content discovery across millions of items; achieved a 2× improvement in retrieval metrics by reimplementing SOTA embedding techniques from recent papers.

Machine Learning Contributor (Contract), Shipd by Datacurve (YC W24)

Awarded monetary prizes for top-tier performance in high-stakes predictive modeling and algorithmic problem-solving challenges.

Developed and optimized ML models under strict competitive constraints in a fast-paced environment.

Research Fellow, AI Safety Camp (AISC)

Working with Ihor Kendiukhov to evaluate scalability and security guarantees of novel control protocol classes for advanced AI systems.

Extending Greenblatt et al.'s framework by designing hierarchical and parallel control topologies to optimize safety-usefulness trade-offs.

Conducting scaling experiments to verify generalization of control guarantees across diverse model capabilities.

Independent Interpretability Researcher, Mentored by David Africa (UK AI Security Institute)

Conducted independent research on mechanistic interpretability, with direct technical guidance from a Research Scientist at the UK AI Safety Institute.

Developed a framework to permanently embed steering vectors into model weights, enabling persistent safety behaviors without inference-time compute overhead.

Research Intern, International Institute of Information Technology, Hyderabad

Researched Universal Semantic Representations and developed techniques for generating coherent natural language sentences from abstract syntactic-semantic structures.

Worked on Controlled Image-to-Text Generation systems for scientific images to ensure accurate, context-aware, domain-specific textual descriptions.

Machine Learning Intern, Co-build.tech

Enhanced query performance by 25% by optimizing document embedding strategies for semantic similarity retrieval.

Automated mapping of legal clause dependencies, improving analysis efficiency by 30%.