Product Data Scientist

github

Aashish Paruvada  • Peerlist

This is an implementation of Image detection using OpenCV Library. This repository contains two files for face detection through both Images and Webcam live face detection. The Haarcascade Frontal Face XML File has been used as a pre-trained model. Haar cascades also have a wide range of pre-trained models that can be directly used to forge out the results from the model.

Basic Face Detection using OpenCV

This is an implementation of Image detection using OpenCV Library and Mediapipe Library. This project automatically detects the nodes on human hands which can later be used for many functions.

Hand Tracking Using Open Computer Vision

HealthSense – End-to-End Health Data Analytics Platform

Stack: Python | Apache Airflow | FastAPI | PostgreSQL | Docker | Kubernetes | GCP | Terraform

Developed a scalable health data analytics platform that fetches, transforms, and stores real-time public health data (e.g., COVID-19 stats from WHO API). Built automated ETL pipelines using Apache Airflow, exposed data services via FastAPI, and containerized the entire stack with Docker. Orchestrated deployments via Kubernetes and infrastructure as code with Terraform on Google Cloud Platform. Implemented monitoring using Prometheus and enabled CORS for seamless frontend-backend integration.

Key Points:
• Automated ingestion and transformation of large-scale health data
• Built modular and reproducible ETL pipelines using Apache Airflow DAGs
• Designed RESTful APIs for health data access using FastAPI
• Deployed containerized services on GCP Kubernetes Engine
• Managed infrastructure using Terraform with cost-aware resource provisioning
• Enabled API monitoring and metrics collection using Prometheus
• Followed best practices in data engineering, cloud DevOps, and API design

Outcome:
The HealthSense platform is a basic Analytics Dashboard. That is mainly built to make use of all the available Infra. This is aimed as a learning challenge.

HealthSense Platform

One day, I was at my father's and we had this game called Housie. The game rules are simple, each one will have to buy at least a ticket (similar to a lottery ticket), and the prize money depends on the sum of money collected by the people purchasing tickets. Once the final sum is counted, it will be allocated to various prizes. There would be a fast-five, 3 Lines, and a housie. 

Once everyone has their tickets, one person would start picking up coins ranging from 1-90 from a box and announce it loud to the players. If the number exists in the ticket, the player would strike it off. The player who gets the first five numbers wins the prize set for first-five, and the player who first strikes all the numbers in the same row wins a line. Every ticket has 3 rows. The first player to get all the numbers to strike off in the same ticket wins a housie.

After long hours of play, I realised that it was tiring for the person who picked and announced the number. So, I came up with this UI using Next.js to make the process easy. We could open this website in the TV and easily play for hours. Visit to know more!

Housie Number Picker

I've built a chatbot that uses RAG implementation to fetch information from the documents or the sources that we mention. I've used OpenAI Embeddings and ChromaDB as the VectorDB to store the Embeddings. To fetch the relevant chunks of information I have used cosine similarity. To not lose context while chunking, I have tried implementing overlapping chunk windows and metadata ids for the chunks. Finally, used n8n for workflow automation and Streamli UI to interact with the bot.

RAG based Chatbot

Reverse image search is a content-based image retrieval (CBIR) query technique that involves providing the CBIR system with a sample image that it will then base its search on in terms of information retrieval, the sample image is very useful in its ways. In particular, reverse image search is characterized by a lack of search terms. This effectively removes the need for a user to guess at keywords or terms that may or may not return a correct result. A reverse image search also allows users to discover content that is related to a specific sample image, the popularity of an image, and discover manipulated versions and derivative works.