ACDC

• Built an Agentic AI product that assists sales teams. Used LangGraph (ReAct with tools), Qdrant, Mem0.
• Fine-tuned Kimi K2 on UAE construction industry data.
• Handled the business ops and customer management.

Infocusp

• Part of a team at Google X working on genomics & agriculture
• Picked genes to be modified for desired traits.
• Used Monte Carlo Dropout & Variational Dropout in NN arch to reduce the effects of network sparsity.
• Introduced & built a pipeline, fully YAML-configured, supporting custom NN arch + all Scikit-Learn models. Hyper-parameter tunable & automated VertexAI deployments.
• Built 7 data pipelines interpreting & transforming raw DNA sequences to trainable features.
• Developed a CV system on Vertex AI with crowd-sourced labelling to automate data collection from live footage.

Infocusp

• Autodesk team: Time-series forecasting models for chemical dosage and power consumption for water treatment plants.
• Bryte team: Wrote AWS infrastructure services to preprocess smart-bed data coming in real-time.
• Legal Sifter team: Fine-tuned BERT architectures to optimize embedding strategies for legal documents.

Schneider Electric

• Researched ML approaches for real-time phishing detection (URL structure & webpage contents).
• Analyzed energy consumption data to identify usage patterns and operational anomalies.

Enlighten Infosystems

• Built an interactive geospatial visualization tool in Python and Bokeh.
• Implemented swarm algorithms for feature selection on genetic data - Artificial Bee Colony, Ant Colony Optimization and Genetic Algorithm.

Bhaskaracharya Institute of Space Applications and Geo-Informatics

• Built 3 plugins to automate GIS workflows in the org.
• 1 plugin reduced manual trimming work of ~5mins per spatial feature to seconds.

Indian Institute of Technology (IIT), Madras

Charotar University of Science and Technology