Software Development Superstream: Generative AI to Improve Software Development

Sam Newman: Introduction (5 minutes) - 8:00am PT | 11:00am ET | 3:00pm UTC

• Sam Newman welcomes you to the Software Development Superstream.

Abhishek Kumar: Unlocking the Future—Generative AI as the Catalyst for Modern Software Innovation (40 minutes) - 8:05am PT | 11:05am ET | 3:05pm UTC

• Generative AI is transforming the landscape of software engineering, reshaping how we approach every stage of the software development lifecycle (SDLC). This groundbreaking technology promises to boost productivity, enhance quality, and speed up time-to-market. Abhishek Kumar explores the current state of generative AI, its diverse applications, and the immense potential it holds for the future of software development.
• Abhishek Kumar is senior director of data science and leads the AI practice at Publicis Sapient. He works with Fortune 500 companies on data and AI strategy and large-scale implementation for recommendation engines, anomaly detection, optimization, and generative AI. He is recognized as a Google Developer Expert-ML and is featured in the 40 Under 40 list of data scientists. Abhishek has authored several popular data science courses and has delivered talks at global data and AI conferences. He’s also a learning facilitator and instructor at Berkeley Haas Executive Education, focusing on the future of technology. He holds a master’s degree from the University of California, Berkeley, and is a recipient of the prestigious Hal Varian MIDS Capstone Award.

Enys Mones: Refactoring Versus Refuctoring—Code Quality in the Al Age (40 minutes) - 8:45am PT | 11:45am ET | 3:45pm UTC

• The emergence of large language models promises unprecedented improvements in software development. Writing code has never been easier or faster, but research shows that the overall time spent on software engineering is unevenly distributed across different mental processes of the engineer. Enys Mones puts AI-assisted code generation in context, highlights the invisible problems that come with using these tools without proper checks and guardrails, and shows how to use AI to obtain software code that you can trust.
• Enys Mones is the lead data scientist at CodeScene. A physicist by training, he pursued data science after obtaining his PhD in theoretical physics. At CodeScene, his main focus areas are designing and curating the company's R&D data lake as well as building statistical models to enhance CodeScene's analytical capabilities. In his spare time, he likes to contribute to open source projects and play music.
• Break (5 minutes)

John Feeney: Using Personalized AI Agents to Speed Up Software Development and Improve Code Quality (Sponsored by Tabnine) (30 minutes) - 9:30am PT | 12:30pm ET | 4:30pm UTC

• AI code assistants are becoming an expected part of every developer’s tool chain. But not everyone is seeing real benefits from them, and concerns persist about the quality of what they generate. John Feeney reveals practical ways to make AI tools more relevant to your specific needs, significantly enhancing developer productivity and code quality. He explains the role of retrieval-augmented generation (RAG) and the importance of using your existing codebase as context and offers guidance on the best way to customize and fine-tune AI coding agents.
• John Feeney is a principal architect within the CTO Office at Tabnine. Prior to helping Tabnine's customers accelerate their software development cycles through the use of AI, he worked with various APM and DevOps tooling vendors to help drive development and operational efficiencies. He is based in Dublin, Ireland.
• This session will be followed by a 30-minute Q&A in a breakout room. Stop by if you have more questions for John.

Ines Montani: Taking LLMs Out of the Black Box—A Practical Guide to Human-in-the-Loop Distillation (40 minutes) - 10:00am PT | 1:00pm ET | 5:00pm UTC

• As the field of natural language processing advances and new ideas develop, we’re seeing more and more ways to use compute efficiently, producing AI systems that are cheaper to run and easier to control. Large language models have enormous potential, but they also challenge existing workflows that require modularity, transparency, and data privacy. Ines Montani shows some practical solutions for using state-of-the-art models in real-world applications and distilling their knowledge into smaller and faster components that you can run and maintain in-house. She shares some real-world case studies and approaches for using large generative models at development time instead of runtime, curating their structured predictions with an efficient human-in-the-loop workflow, and distilling task-specific components that run cheaply, privately, and reliably, and that you can compose into larger NLP systems.
• Ines Montani is a developer who specializes in tools for AI and NLP technology. She’s the cofounder and CEO of Explosion and a core developer of spaCy, a popular open source library for natural language processing in Python; and Prodigy, a modern annotation tool for creating training data for machine learning models.
• Break (5 minutes)

Pamela Fox: Building a RAG App to Chat with Your Data (40 minutes) - 10:45am PT | 1:45pm ET | 5:45pm UTC

• One of the most popular use cases for generative AI is retrieval-augmented generation (RAG), a technique that feeds context to a large language model so that it can answer questions according to sources. Thousands of developers have deployed RAG applications to chat with their internal enterprise data (like HR or support) and external public data (like government or retail). Pamela Fox provides an overview and best practices for building a RAG app.
• Pamela Fox is a principal cloud advocate in Python at Microsoft. Previously, she was a lecturer for UC Berkeley, the creator of the computer programming curriculum for Khan Academy, an early engineer at Coursera, and a developer advocate at Google.

Sam Newman: Closing Remarks (5 minutes) - 11:25am PT | 2:25pm ET | 6:25pm UTC

• Sam Newman closes out today’s event.