Hi, I’m Kiriti.

On the core Codex team at OpenAI, I worked across product development, agent behavior and reliability, and production infrastructure. Examples include the Codex Web launch, Slack integration, Code Review, and GPU capacity and fleet management.

I co-created ChatGPT Sites and co-led its development and launch.

What I learned building Codex

As a founding ML infrastructure engineer, I helped build Kumo's predictive AI platform, which uses graph neural networks to learn from relationships across enterprise tables. I helped scale it from a single node to distributed infrastructure training hundreds of models a day.

I built compute infrastructure handling hundreds of terabytes and a RocksDB feature store for GNN training. I led native Snowflake and Databricks integrations, including running training and inference inside customers' Snowflake accounts with Snowpark Container Services.

My work at Google spanned distributed databases and ads infrastructure, including machine learning use cases in Google Ads.

I contributed to a distributed data store serving 3 million requests per second across regions. My work included RDMA caching and fault-tolerant processing to improve performance and support workloads at scale.

On the data lake team, I built a Spark–Salesforce connector and data pipeline features for checkpointing, query replay, SLAs, and alerting.

In Samsung's Big Data Lab, I built infrastructure for the Smart TV recommendation ML platform. My work on the data pipelines team covered Spark pipelines processing terabytes of data each day and real-time recommendations with Spark Streaming.