Work on large-scale Retrieval-Augmented Generation (RAG) and agentic systems, designing modular workflows and optimizing inference performance across LLM-based applications. Design and deploy agentic workflows with LangChain and LlamaIndex, and implement hybrid search pipelines using Qdrant. Optimize LLM training and inference with custom CUDA and Triton kernels, reducing latency and compute costs. Build backend APIs with Python and FastAPI to serve embeddings, RAG pipelines, and agent execution, integrating custom PyTorch extensions for high-performance GPU workloads.

Developed a Retrieval Augmented Generation (RAG) platform combining large language models (LLMs) with semantic search. Built an API using FastAPI with endpoints for embedding generation, data ingestion, and queries to a Qdrant vector database. Implemented hybrid search using dense, sparse, and late interaction embeddings for re-ranking, enabling efficient retrieval over a dataset of 70+ million documents.

Created technical content such as tutorials, videos, and documentation about semantic search, vector search, and retrieval augmented generation (RAG) using LangChain, LlamaIndex, Qdrant, OpenAI, and various LLMs. Developed demos and sample projects leading to a YouTube channel growth of almost 400%. Served as internal consultant solving integration issues with Qdrant for large-scale retrieval systems in production environments.

Specialized in cloud infrastructure using Terraform, Ansible, Docker, and Kubernetes (EKS) to create robust and scalable systems. Collaborated on migrating Ruby on Rails applications from bare-metal infrastructure to the cloud. Focused on application observability with Loki, Grafana, Datadog, and eBPF. Collaborated with data science team building machine learning pipelines using PyTorch.

Refined data quality, automated machine learning pipelines, and optimized model development lifecycle. Built and deployed complex ML and Deep Learning models using Python, Scikit-Learn, and TensorFlow for predictive analysis. Developed a computer vision model identifying nutritional value of food using a CNN built with PyTorch. Built reliable ML workflows in production using Airflow, Kubeflow, and Weights & Biases.

Managed on-call incident response ensuring correct functioning of large Elasticsearch clusters and deployment of new regions for Elastic Cloud, ensuring high availability in AWS, Azure, and GCP multi-regional and multi-cloud environments. Developed CLI tooling in Golang for Kubernetes cluster maintenance. Used Ansible, Terraform, Python, and Scala as an SRE.

Worked with AWS, GCP, Docker, Kubernetes, Ansible, Terraform, Prometheus, and Grafana.

Developed software using PHP, Java, Ruby on Rails, MySQL, Postgres, LXC, and Docker.

Managed Linux systems, iptables, proxies, networks, Bash, and Perl scripting.