Amit Dutta

Software Engineer

Meta • Fulltime

Sep 2025 - Present • 8 mos

• Built a fault injection observability dashboard for an infrastructure experimentation platform used by 5,000+ engineers running 10,000+ experiments daily for chaos engineering experiments. • Developed an AI-assisted debugging feature using LLM agents to analyze infrastructure failures and deliver actionable fix suggestions directly in engineers’ IDEs, reducing average debug time from 30 minutes to 5 minutes (6x faster) for failed experiments.

Software Cloud Engineer

Google • Fulltime

Jul 2022 - Jul 2025 • 3 yrs

• Developed an AI-powered solution to automate and streamline nurse shift handovers at Manipal Hospitals by synthesizing notes into ISBAR format, resulting in a 70% time reduction. • Developed Navi, a LLM playground (Gemini, Codestral and OpenAI models) with customizable parameters (temperature, tokens, etc.) and chat history adopted by 100+ beta users with 5s response times. • Designed a dynamic webapp for more than 500 Albertsons merchandisers to build and manage product pathway widgets (drag and drop, boost / bury, timed activation), significantly improving merchandising efficiency and product display control in 2000+ stores. • Build an AI-powered interactive interface enabling over 300 technicians to query vehicle manuals for repair and maintenance procedures with up to 90% accuracy, including outcome summaries with links, page references, and relevant images/diagrams from over 1,000 manuals. • Developed a Text-to-JSON and JSON-to-image generation service for Betmakers processing over 100 racecourse data entries to produce more than 300 images. Exceeded a 90% accuracy target, achieving 93.8% success with an average image generation time of 20 seconds, enabling real-time data visualization for customer production use.