Aug 2023 – Present · 2 yrs 9 mos

• Led the development of a novel platform to evaluate LLM-generated code within Intuit's specific context, addressing the inadequacy of traditional metrics (BLEU, ROUGE-L, CodeBLEU, CodeBERT) and existing research (e.g., ICE score) for accurate assessment.
• Engineered this evaluation platform with an indigenous algorithm, now integral to the capability onboarding pipeline for integrating Intuit code with LLM-assisted development. Successfully reduced human evaluation effort by 1 hour per capability, significantly streamlining the onboarding process.
• Utilized AI agents to automate the transformation of SDKs into curated "golden repositories" of examples, essential for effective Retrieval Augmented Generation (RAG) ingestion for Qodo Gen
• Spearheaded the deployment of Qodo Merge and Qodo Gen, incorporating rigorous performance testing, Failure Mode and Effects Analysis (FMEA), and operational excellence principles.
• Achieved a 31% increase in PR merge velocity and a 20 increase in PR creation velocity, significantly boosting developer efficiency.
• Spearheaded initiatives to enhance platform stability by implementing robust measures to prevent cluster downtime caused by excessive proxy connections, utilizing Pulsar brokers for real-time connection management.

Feb 2021 – Aug 2023 · 2 yrs 6 mos

• A key member of the messaging team which operates a highly distributed, multi-cluster, multi-region messaging platform to serve the queuing use-cases of its applications and service.
• Initiated the adoption of Pulsar for the next generation platform and adapted it for Intuit specific requirements.
• Key contributor to a library which helps customers seamlessly migrate from activeMQ to Pulsar

Aug 2019 – Feb 2021 · 1 yr 6 mos

• Built an automation to replay messages in ActiveMQ queues responsible for store percentage reducing the daily overhead of manually replaying messages.
• Built a service to provide intrinsic metrics very essential for oncall such as finding stuck consumers and finding oldest message for a queue to reduce developer daily effort .