Building production-grade AI systems that scale

I work at the intersection of Machine Learning, Distributed Systems, and Generative AI, focusing on turning research ideas into reliable production platforms.

Currently working as a Data Science Manager, while staying deeply hands-on with:

• AI system architecture
• model development
• scalable ML infrastructure
• production deployment

• Retrieval Augmented Generation (RAG)
• LLM orchestration pipelines
• Prompt engineering & evaluation
• AI agents and tool-based workflows

• Scalable model training pipelines
• Feature engineering at scale
• Experiment tracking frameworks
• Production MLOps workflows

• PySpark based data pipelines
• Streaming architectures
• Data lakehouse systems
• Large-scale data processing

• FastAPI model serving
• GPU inference optimization
• Scalable microservices
• Containerized deployments

• Designing end-to-end GenAI platforms
• Building LLM inference systems on Linux/Ubuntu
• Exploring Snap-based AI model deployment
• Architecting scalable ML platforms on cloud

Enterprise analytics platform that automates weekly and monthly executive business reviews using LLM-generated insights over TB-scale transactional data.

flowchart LR
    A[Transactional Data Sources] --> B[Data Lake / Warehouse]
    B --> C[OLAP Analytics Pipeline]
    C --> D[Feature Engineering Layer]
    D --> E[LLM Insight Generation]
    E --> F[FastAPI Service Layer]
    F --> G[Executive Dashboards / Reports]

Highlights

• OLAP-first analytics pipeline for high-volume business reporting
• Feature engineering for consistent enterprise metrics
• LLM-based narrative insight generation with cost/latency optimization
• FastAPI architecture separating analytics, services, and AI inference
• Governance via caching, rate limiting, and budget controls

➡️ Full case study

Production geospatial ML system supporting new retail site selection and 3-year sales forecasting for a $10B+ retailer with 2,500+ locations.

flowchart LR
    A[Geospatial Datasets] --> B[Databricks + Spark Feature Engineering]
    B --> C[Trade Area Modeling]
    C --> D[XGBoost Forecasting Models]
    D --> E[Explainability Layer SHAP]
    E --> F[MLflow Model Registry]
    F --> G[Serverless Model Serving]
    G --> H[Real Estate Decision Platform]

Highlights

• Large-scale geospatial feature engineering using Spark and Databricks
• Dual trade-area modeling using drive-time isochrones and radial distance
• Cold-start modeling via clustering-based “statistical twins”
• Multi-category XGBoost forecasting architecture
• Explainability-first design using SHAP for executive trust
• Production deployment using MLflow, Delta Lake, and serverless serving

➡️ Full case study

Python • SQL

PySpark • Pandas • Scikit-learn • TensorFlow • PyTorch

LLM APIs • RAG frameworks • Vector Search

Docker • Kubernetes • FastAPI • Linux / Ubuntu

GCP • Distributed ML Platforms

Outside work:

🏍 Motorcycling through mountain roads
🧠 Exploring emerging AI technologies
👥 Building high-performance engineering teams

Opportunities to build large-scale AI platforms and Generative AI systems.

Roles

• Senior AI / ML Engineer
• AI Architect
• Applied AI Leadership

Locations

• India
• Middle East

Great AI systems are not just models.

They are well-designed systems built with:

• strong data foundations
• reliable infrastructure
• disciplined engineering