# Ajay Benno > Infrastructure @ Assembled Location: New York, New York, United States Profile: https://flows.cv/ajaybenno ## Work Experience ### Senior Software Engineer - AI Agents @ Assembled Jan 2025 – Present Working on our Chat AI team - Decreased chat latency by 25% by identifying LLM inference latency due to bloated responses - Added the ability to have cost attribution for our in house model provider service. This allows us to see spend across our different models, products and vendors and identify cost improvements - Improved our agentic retrieval pipeline by parallelizing the rewriting step. This allows us to better target the source systems and write more optimized queries for each data store. ### Senior Software Engineer - Infrastructure @ Assembled Jan 2024 – Present | New York, NY - Rearchitected and scaled our Postgres layer to reduce daily outages due to connection issues. Implemented a highly available pgbouncer setup in Kubernetes with a passive failover to handle regional failures. Has allowed us to greatly scale our read and write volume to our primary instance with no incidents. - Eliminated single points of failures for our Postgres read replicas and migrated us to DNS for our load balancing. This led to 50% faster p95 and reduced costs due to reduced cross region data transfer. - Championed and introduced distributed tracing to our stack, instrumenting all services to give all teams visibility into transactions. Did this by introducing opentelemetry as the standard framework. - Managed datadog spend, implemented multiple cost reduction improvements around custom metrics and APM. - Reduced Snowflake costs by 50% by analyzing query patterns and implementing clustering keys to reduce warehouse usage. ### Engineering Manager @ Yext Jan 2023 – Jan 2024 Engineering Team Lead of Analytics * Led the development and launch of a Universal Analytics Events API, capable of ingesting thousands of user events per second. This enhancement significantly improved error reporting and increased developer friendliness. It enabled new teams to launch real-time analytics for their products within days, a substantial improvement from the weeks required with the old system. * Led the development of SLOs (Service Level Objectives) for our core service, helping to alert teams when customers faced data delays. * Migrated multiple daily batch jobs to airflow, simplifying the data model and greatly reducing cost on our warehouse spend. * Facilitated bi-weekly scrum practices with planning, estimation, and retrospectives. Focused on singular sprint goals and driving the team to work on a single project at once, which significantly improved the overall team velocity. * Implemented new team practices to manage alerting and maintain consistent oversight of our system health. This resulted in a more manageable overall volume of alerts and a significantly better signal-to-noise ratio. ### Senior Software Engineer @ Yext Jan 2023 – Jan 2023 ### Software Engineer @ Yext Jan 2019 – Jan 2023 | New York, New York •Prototyped KafkaConnect for the company; Currently being used to stream messages from Amazon SQS to Kafka. Handling over 40 million messages per day. •Owned Intelligent Search Tracker, a large scale data scraping service. Interfaced with third party vendors to handle data integrity issues and increase reliability. •Designed a system to automatically propagate data delays from source data systems and surface that to customers. •Deployed tracing system for our deployment pipeline to help aggregate deployment issues at scale. ### Software Engineering Intern @ Yext Jan 2019 – Jan 2019 | New York, New York Analytics Team ### Teaching Assistant - Database Systems @ Carnegie Mellon University Jan 2018 – Jan 2018 | Pittsburgh, PA Teaching Assistant for Database Systems(15-445). ### Software Engineering Intern - Performance Engineering @ Redfin Jan 2018 – Jan 2018 | San Francisco Bay Area Integrated Google Lighthouse into internal performance tests; Lighthouse provided a new lens on performance and is used to catch unintended regressions. Engineered an efficient method of tracking and logging non-responsive interactions across the website; Data will help identify pages that are key sources of customer frustration ### Undergraduate Researcher @ Carnegie Mellon University Jan 2017 – Jan 2018 | Pittsburgh, PA Working with off-the-shelf WiFi cards to classify the material of an environmental object (eg. wood, metal, human) using its physical properties, and to determine what angle the object is at. Helping to run experiments, write code to read the signals, and research new approaches. Since WiFi can propagate around corners and through walls, possible use cases for this technology include autonomous vehicles and searching disaster sites. ### Software Engineering Intern, Marketing and Decisioning @ Capital One Jan 2017 – Jan 2017 | San Francisco, California Implemented a microservice to aggregate data and pipeline it into a model to predict credit card application fraud. Setup CircleCi pipelines which built docker containers for easy deployment. Wrote Cloudformation scripts to automatically configure Amazon EC2 instances. Worked on adding H20.ai support to clipper, an open source prediction serving system. ### Software Engineering Intern - Emerging Technology Center @ Software Engineering Institute | Carnegie Mellon University Jan 2016 – Jan 2017 | Greater Pittsburgh Area Used a corrective gradient refinement algorithm to localize a robot in a physical space. Worked on an application of CGR localization which could autonomously move the robot around in the space by clicking on a map. Contributed to Micro Expression project by switching the classification model to a support vector machine. ### Software Engineering/Machine Learning Intern @ Decisive Analytics Corporation Jan 2016 – Jan 2016 | Arlington, Virginia Worked as a software engineering intern within the machine learning division at Decisive Analytics. Developed a model to classify the popularity of YouTube videos and predict if they would go viral using various machine learning and natural language processing techniques. Generated user profiles by implementing latent dirichlet allocation to run on a users' YouTube interactions. ### Software Engineering Summit @ Capital One Jan 2016 – Jan 2016 | Mclean, Virginia One of 30+ students from 300+ applicants to be selected to attend a one week long Software Engineering Summit in Capital One's corporate headquarters. ### Tutor @ CodeHS Jan 2015 – Jan 2016 Graded problems that were submitted by students. Answered questions about computer science and explained it in easy to understand terms. ## Education ### Bachelor of Science (B.S.) in Electrical and Computer Engineering Carnegie Mellon University Jan 2015 – Jan 2019 ### Methacton High School Jan 2011 – Jan 2015 ## Contact & Social - LinkedIn: https://linkedin.com/in/ajaybenno --- Source: https://flows.cv/ajaybenno JSON Resume: https://flows.cv/ajaybenno/resume.json Last updated: 2026-04-01