I'm an AI/ML Engineer working across deep learning, intelligent systems, data engineering, and applied AI research. I build end-to-end systems — from data pipelines and experimentation to model training, evaluation, APIs, and production deployment.
flowchart LR
A[📦 Data] --> B[🧬 Representation]
B --> C[🤖 Model]
C --> D[🔬 Experimentation]
D --> E[🚀 Deployment]
style A fill:#0F2027,stroke:#36BCF7,color:#fff
style B fill:#1a3540,stroke:#36BCF7,color:#fff
style C fill:#203A43,stroke:#36BCF7,color:#fff
style D fill:#264a56,stroke:#36BCF7,color:#fff
style E fill:#2C5364,stroke:#36BCF7,color:#fff
Current interests:
GNNs·Transformers·Representation Learning·Spatiotemporal ML·Scientific ML·LLM Systems
| 🎯 Role | 🏛️ Organization | 🔍 Focus |
| ML Engineer | IIT Delhi | GNNs · Spatiotemporal Deep Learning · ML Research |
| AI Engineer | AI CoE, IIT Gandhinagar | Deep Learning · AI Systems · HPC/GPU Deployment |
| Sr. Project Technical Assistant | IIT Bombay | ML Pipelines · Time-Series · Large-Scale Data |
| Data Science Intern | IIM Bangalore | Data Science · RAG · Decision-Support Systems |
|
Adaptive RAG with query decomposition, knowledge graphs, multiple reasoning strategies, and conversational memory.
|
AI-powered search & recommendation using asynchronous APIs and locally served LLMs.
|
|
Graph + temporal deep-learning architectures for modelling complex spatial systems.
|
Open-source Python package for air-quality data cleaning, standardization & processing. pip install airpy-tool |
|
🐍 Languages & Data
🤖 AI · ML · Deep Learning
💬 LLMs · RAG · Agentic AI
|
⚙️ Backend · MLOps · Cloud
🔬 Experimentation & Research
🌍 Geospatial · Automation
|
- 📄 Substantially underestimated health burden of Indian road transportation air pollution — Nature Sustainability (Under Consideration)
- 📄 Assessing Performance of the Jal Jeevan Mission Using a Geospatial Decision Support System — Center for Public Policy · IIM Bangalore
- 📄 Central Bank Digital Currency (CBDC) and Application Development —
DOI: 10.22214/ijraset.2023.57703
🕸️ Graph Neural Networks
Learning over graph-structured data — message passing, attention, and pooling for relational and spatial systems. Core to my spatiotemporal modelling on networks like river basins and sensor grids.
🧬 Representation Learning
Discovering compact, meaningful embeddings from raw signals — self-supervision, contrastive objectives, and manifold structure that make downstream tasks easier and more transferable.
🔀 Transformers & Vision Transformers
Attention-based architectures for sequences and images — from language modelling to patch-based ViTs for segmentation and remote-sensing imagery.
🌀 State-Space Models
Efficient long-sequence modelling (S4 / Mamba-style) — linear-time alternatives to attention for very long temporal and spatial dependencies.
🔭 Scientific Machine Learning
Blending physical laws with data-driven models — physics-informed losses, differentiable simulators, and neural operators for scientifically-consistent predictions.
🛰️ Geospatial Foundation Models
Large pretrained models for satellite and Earth-observation data — reusable representations for classification, segmentation, and change detection across regions.
🤝 Agentic AI Systems
LLM-driven agents that plan, use tools, and coordinate — orchestration with LangGraph/MCP, memory, and multi-step reasoning for autonomous workflows.
AI · Machine Learning · Deep Learning · Intelligent Systems