saran M

Designed and deployed automated data pipelines in Python to collect and process large-scale analytics from multiple APIs. Built RESTful APIs and backend services for internal dashboards and data analysis systems. Developed format conversion pipelines for content publishing (Apple News Format). Developed document digitization pipelines using Azure Vision API and LLMs (Ollama) to extract and structure information from unstructured sources. Implemented AI-powered content extraction and classification systems integrated into enterprise data workflows. Created Python-based micro-services that sync and automate task data between third-party SaaS platforms. Deployed and maintained services on AWS EC2, using Docker, cron jobs, and SSH-based automation. Collaborated with data and product teams to integrate AI features into production environments with high reliability.

Designed retrieval-augmented generation (RAG) systems and utilized vector databases for efficient storage and optimizing information retrieval processes. Handled end-to-end deployments on AWS EC2 using Nginx, cron, systemctl, Forever and SSH, ensuring uptime and performance optimization. Proficient in LangChain for developing applications using large language models (LLMs) and OpenAI models. Developed robust web scraping solutions using Python libraries like Beautiful Soup and Requests, enhancing data processing workflows with Langchain for seamless integration of language models. Utilized Google Cloud’s Vertex AI platform to build and deploy machine learning models, leveraging its unified interface for the entire ML workflow from data preparation to model deployment. Developed a web scraping project utilizing Large Language Models (LLMs) and Ollama, automating data extraction and processing from various websites. Gained hands-on experience in connecting React applications with APIs, enhancing data retrieval and user interaction. Converted a React application into a standalone executable file. Designed and implemented a Windows-based background task using Detectron2, Tesseract-OCR