Aman Nipun Shah

Implemented a peer-to-peer engagement feature that allows users to endorse peers for specific skills. Endorsements are visible to both users and their reporting managers, aiding in assessing role eligibility. I also developed a recommendation engine to highlight relevant skills for endorsement, optimizing the process to reduce job completion time from 82 minutes to just four minutes. This solution efficiently processes around 700 – 800 million rows using Node.js, TypeScript, Amazon ECS, and PostgreSQL, incorporating multiple checks and enhancements.

Created a cache management system at Cimpress using Amazon SQS, Node.js, and AWS Lambda to efficiently clear product pricing caches during bulk updates. When the admin team updated prices, notifications were sent to SQS, triggering multiple Lambdas in parallel at the database's maximum absorption rate. Each Lambda independently cleared caches for various 3rd-party services and the database itself, reducing cache invalidation time by 70% and enabling dynamic pricing updates.

Integrated the Stripe API using a webhook model in Scala. I implemented a redirect to the Stripe payment page upon clicking the "Pay Now" button, configuring the system to return users to the appropriate success or failure page after payment. I also managed webhooks from Stripe to display the payment status and amount, ensuring a seamless user experience.

Developed integration with Autotask as part of the back-end team at Zomentum to export real-time changes from the portal. I created a trigger for entity updates on our portal that sends messages to Amazon SNS, notifying an Amazon SQS queue. A Lambda function was implemented to listen to this queue and update entities in Autotask via their REST APIs. I used Scala for back-end development to ensure efficient processing and integration.

Developed an application that automates the process of obtaining rail concessions for college students. The platform features dedicated modules for students, colleges, and railway authorities, streamlining communication and management. Students receive email updates on their application status, enhancing transparency and engagement. I utilized Node.js for back-end development and MySQL as the database to ensure robust data handling and performance.

Built an AI-powered call analysis platform to generate insights from customer interactions such as sales, onboarding, and introductory calls. Developed a scalable orchestration service that integrates multiple AI models, including OpenAI, Claude, and Gemini, allowing users to choose the preferred model for analysis. The system used a multi-layer architecture managed through SQS queues, where the initial layer classified the call type and routed it to specialized processing pipelines. Each call category used tailored prompts to generate detailed insights, summaries, and actionable recommendations. Integrated a conversational chatbot interface that enables users to ask questions about their call data and receive contextual insights. The platform also generated periodic performance summaries to help teams evaluate call agent effectiveness, identify trends, and improve customer engagement. The system was designed for scalability and efficiency, enabling automated processing and analysis of large volumes of call data while delivering meaningful business intelligence.

Developed an AI-powered DIY reporting and dashboard platform that allows users to generate insights using natural language prompts. Built a model-agnostic pipeline with LangChain to convert user prompts into SQL queries executed on existing Parquet datasets. The system dynamically retrieves data and uses AI models to determine the most suitable visualizations, automatically rendering interactive dashboards. Back-end services were implemented in TypeScript using the OpenAI SDK, while the front end was built with React for a responsive user experience. Additionally, implemented a high-performance ETL pipeline in Go to process Snowflake data and write it to Parquet files, enabling runtime joins for report generation. This optimized workflow handled large datasets, with reports containing 10+ million rows and sizes exceeding 6GB, achieving a 10× cost reduction and 22× faster performance. The platform empowers users to self-serve insights efficiently by combining AI-driven query generation, visualization recommendations, and scalable data processing into a seamless reporting experience.

Developed a high-performance reporting and dashboard system in Go, leveraging Snowflake and Parquet files for scalable analytics. Built ETL pipelines to store data in Snowflake, convert it to Parquet, and perform runtime joins for dynamic report generation. The system efficiently handles large datasets, with reports exceeding 10 million rows and 6GB in size, while achieving a 10× reduction in cost and 22× improvement in processing speed. Implemented automated workflows to generate dashboards and reports on demand, enabling users to derive actionable insights quickly. The platform integrates back-end Go services with optimized data handling, ensuring minimal latency and high reliability for large-scale analytics. By combining ETL automation, Parquet-based storage, and runtime data joins, the solution delivers fast, cost-efficient, and scalable reporting for enterprise-level datasets.