Senior Full Stack Engineer | Mentor | Solopreneur | Toptal

github

BHAVYA BHUSHAN  • Peerlist

This project involved creating a web application that uses OpenAI's ChatGPT to improve how the Business team understands client technical needs. The application asks clients targeted questions, analyzes their responses, and generates detailed talent personas. These personas help the team match clients with the most suitable professionals, ensuring successful project outcomes.

Built with React, TypeScript, Express, Prisma, and GCP, this solution offers several key benefits:

• Efficient client interaction: Gathers crucial information through dynamic, AI-driven questions.
• Improved talent matching: Creates data-driven personas for more effective talent allocation.
• Increased team efficiency: Automates key processes, freeing up time for higher-value tasks.
• Enhanced client satisfaction: Leads to better project outcomes and stronger client relationships.

AI-powered Persona Generator for Enhanced Client Understandi

Reference : http://yann.lecun.com/exdb/mnist/index.html

MNIST ("Modified National Institute of Standards and Technology") is the de facto “hello world” dataset of computer vision. Since its release in 1999, this classic dataset of handwritten images has served as the basis for benchmarking classification algorithms. As new machine learning techniques emerge, MNIST remains a reliable resource for researchers and learners alike. 
In this repo, the goal is to correctly identify digits from a dataset of tens of thousands of handwritten images.

Tools used : Python2.7, Keras, Sklearn, CNN, Matplotlib

Result : Predicted on testing-dataset with 99.27% accuracy

Digit-Recognizer

This project involved solving the issue of fragmented data in BigQuery and various sheets. In addition, siloed teams and authorization limitations hindered transparency. I tackled these problems with a multifaceted approach, collaborative efforts, and a creative solution. Open communication across data teams pinpointed data sources. My creative solution was a Google Apps Script proof-of-concept dashboard to visualize real-time performance data.

Leveraging my developer skills, I built the dashboard with an ETL pipeline for automated data updates, showcasing the potential for improved performance transparency. While the solution validated the concept, limitations like scalability and role-based access became evident. Demonstrating project ownership, I championed a full-stack web application for a long-term solution.

The final web application offers a robust solution, secure and scalable access to performance data, and role-based access control. This empowers teams and managers with personalized insights, fostering informed decision-making and improved organizational performance.

Performance Transparency Dashboard

This project addressed the need to streamline and improve the quality assurance process for various call types, including technical interviews and matching calls. The resulting web application provides QA teams with automated, dynamic call assignments based on their availability, business vertical weightage, and monthly targets.

A key feature of the application is the integrated display of call recordings from platforms like BlueJeans and Zoom within a single screen. Additionally, the application provides detailed, call-specific scoring mechanisms, enabling comprehensive and standardized quality evaluations.

Developed with an ETL pipeline from BigQuery to PostgreSQL, the application interacts with the database to manage assignments and scoring data. Built using React, TypeScript, and Express, it's deployed on GCP using serverless technologies like Cloud Run and Cloud Build.

Impact:
• Automated call assignments for balanced workload distribution.
• Centralized call recordings for increased efficiency.
• Standardized scoring for objective quality evaluations.
• Improved QA process, leading to increased productivity.

QA Workflow Optimization and Efficiency Enhancement

QuickLaunch is a web-based application that assists solopreneurs, indie hackers, and individuals with innovative ideas in transitioning from concept to product. The application leverages AI-assisted tools and deep market analysis to validate ideas, establish product-market fit, and optimize the development process, catering to users with varying skill levels worldwide. Features: Product Creation: From Idea to Product Detailing - DONE<ul><li> AI Generated Realtime ques/answer to get better understanding of the product as per user’s knowledge</li><li>Detailed Product Creation, understanding bottlenecks, competitors, Features, etc</li><li>Extensive Market Research, to identify PMF, Competitor ANalysis, Marketing Strategies in the Product Analysis Report</li><li>Idea Validation and Sentiment Analysis of the idea on multiple sub-reddits to understand consumer base</li></ul>Product Development: Product Detailing to Actual Product - COMIING UP NEXT<ul><li>AI Generated Ques/Ans to understand the flow of the application and generating user stories, identify external dependencies, to get a complete understanding of how the application would be going to work</li><li>AI Agents taking up the role of Project Manager, Solution Architect, UI/UX Designer to generate a detailed (to minute details) User Stories, Tasks, Sub-Tasks,etc to move straight to development</li><li>Possible Integrations with Project Management tools like Jira, Asana, Trello etcI founded and built this product from the ground up, overseeing all aspects of development, planning, and marketing from start to finish. I used React, Vite, TypeScript, Supabase, Edge Functions, Cloud Function, CrewAI agents, Python, and Pydantic.</li></ul>

QuickLaunch - AI Assisted Idea to MVP Journey

Developed a user-friendly web application (Angular) to streamline CRISPR screening data analysis.

This innovative solution tackles challenges faced by scientists.

CHALLENGES
• Limited accessibility: Script-based tools like MAGeCK require bioinformatics expertise. My application offers a user-friendly interface, democratizing data processing.
• Computational bottlenecks: Traditional analysis takes hours on local machines. I implemented an FTP-based solution for large datasets, enabling efficient back-end processing.
• Data visualization limitations: Text outputs hinder interpretation. I designed a comprehensive dashboard for visualizing comparisons and identifying gene targets.
• Collaboration barriers: Sharing data often involves external tools. The platform facilitates collaboration through project sharing, notes, and comments.

MY SKILLS
• Problem-solving: Identified key limitations of existing tools and developed a user-centric solution.
• Development: Built a robust web application with Angular and implemented efficient data handling mechanisms.
• Innovation: Introduced novel features like state management for upload failures and a collaborative analysis environment.