Chancellor's Scholar and MSc AI & ML candidate at the University of Birmingham specializing in end-to-end production systems. Experienced in architecting complex multi-agent orchestrations, isolated data loops, and deploying predictive machine learning workflows under highly constrained real-world environments.
Designed and built an intelligent agentic system leveraging Google ADK, Vertex AI Vector Search, and Gemini 2.5 Pro to orchestrate parallel pipelines across 10 semantic agroforestry paths. Implemented an interactive state-preserving chat loop inside a programmatic Mesop interface to handle live network plan regeneration without context fragmentation.
Owned the digital twin predictive framework modeling microclimate dynamics for high-altitude artificial glaciers. Deployed edge telemetry streams on physical Raspberry Pi/ESP-32 nodes via secure Cloudflare Tunnels and implemented an LSTM multivariate Autoencoder/Isolation Forest framework that replaced rule-based thresholds to achieve a 40% efficiency gain.
Architecting a phase-by-phase multi-agent loop to translate unstructured engineering blueprints into citation-backed intelligence engines, replacing complex manual technical reviews with deterministic automated auditing lookups. Integrating continuous evaluation pipelines via Ragas to rigorously track semantic overlap and precision metrics.
Developed a localized PyTorch image captioning pipeline and programmatically translated the Flickr8k dataset into 20+ regional Indian languages (30M+ tokenized characters) to resolve data scarcity bottlenecks for native multi-lingual vision architectures.
Recipient of the global Chancellor's Merit Scholarship (awarded to 1 of 10 global recipients). Achieved Distinction status with top marks in Neural Computation (84) and NLP (83).