cv

Lead AI Research Engineer

VFS Global

Leading AI team building production intelligent document processing systems processing 180K+ documents daily across 110+ countries.

• Built production-scale IDP system achieving 99.1% accuracy, reducing visa review time from 60+ minutes to sub-5 minutes
• Developed epistemic uncertainty quantification system with dual-model validation reducing hallucination rates
• Integrated semantic entropy pipeline for distinguishing linguistic variance from epistemic model failure
• Optimized ML model inference by 40% through quantization and vLLM deployment achieving 20 tokens/second
• Led technical team of 8+ engineers

Machine Learning Research Intern

Adani Digital Labs

Developed ML systems for content automation, computer vision, and multimodal search.

• Delivered content automation pipeline using OpenAI APIs generating SEO-optimized content across 150K+ pages, achieving $3.7M cost savings
• Built computer vision classifier using DINOv2 embeddings achieving 93% precision on 500+ image dataset
• Implemented multimodal search engine using CLIP and GPT-4 Vision for product discovery
• Developed self-supervised representation learning for few-shot clustering in low-data regimes

B.Tech

Vellore Institute of Technology, Bhopal

Computer Science and Engineering

B.Sc.

Indian Institute of Technology, Madras

Programming and Data Science

Repair of Thought: Advancing Automated Program Repair through a Dual-Model Reasoning Framework

Submitted for Review

Proposed function-level APR framework achieving 83.1% plausible repair rate on Defects4J. Decoupled reasoning from synthesis, validating repairs via control-flow symbolic analysis and semantic verification.

Image Aesthetics Quantification

Built CLIP-based aesthetics scorer for hotel media ranking with domain adaptation and search integration.

• Precomputed averaged positive/negative vectors reducing per-image comparisons from ~2,000 to 2
• Delivered Streamlit + Colab with domain adaptation capabilities