Building Intelligent Systems at Scale
Architecting production ML pipelines, RAG systems, and MLOps infrastructure that power real-world AI applications serving thousands of users.
Expertise & Focus
AI/ML Engineering
RAG systems, AI agents, LLMs, prompt engineering, fine-tuning, and inference optimization for production environments
MLOps & Automation
End-to-end ML pipelines, CI/CD automation, model monitoring, and deployment orchestration at enterprise scale
Cloud Infrastructure
AWS, Azure, GCP services for ML workloads, infrastructure as code, and scalable cloud-native architectures
Data Engineering
Data pipelines, feature engineering, vector databases, dataset optimization, and distributed data processing
Engineering Philosophy
I believe in building ML systems that are not just accurate, but reliable, scalable, and maintainable. My approach combines rigorous engineering practices with practical AI implementation—cutting deployment times from hours to minutes while maintaining 99%+ uptime. Every system I build is designed to evolve, monitor itself, and serve real users in production.
Featured Work
Medical LLM Chatbot
Built and deployed a production-grade medical chatbot using LangChain, Groq LLM, and Tavily search with complete MLOps infrastructure. Implemented CI/CD pipeline with Jenkins, Docker, SonarQube, and AWS ECS Fargate, reducing deployment time from 2 hours to 15 minutes.
- LangChain
- Jenkins
- Docker
- AWS ECS
- FastAPI
- Prometheus
Multilingual Document Chatbot
Led development of enterprise RAG-based chatbot serving 2,000+ employees across 8 data centers. Architected document retrieval system supporting 4 languages with vector database optimization, improving retrieval efficiency by 95% through advanced chunking strategies and embedding optimization.
- RAG
- Vector DB
- Azure
- Kubernetes
- Python
- NLP
Cancer Survival Prediction System
Orchestrated complete ML pipeline for colorectal cancer survival prediction on 167K+ patient records using Kubeflow on Minikube. Containerized data processing and model training components, reducing pipeline execution from 3 hours to 20 minutes. Integrated MLflow with DagsHub for experiment tracking and deployed Flask API for clinical decision support.
- Kubeflow
- MLflow
- Docker
- Flask
- Scikit-learn
- DagsHub