Friday Hacks #248, October 27
Posted on by Parth Gujar
Food 🍕 will be served!
If you are unable to come join us physically, you are welcome to join us online!
1) Taming open code LLMs for SQL generation and bug fixing
This talk will include a hands-on session that will guide participants through the exciting world of fine-tuning Large Language Models (LLMs) using state-of-the-art techniques such as PEFT and QLoRA.
During this session, participants will discover how to harness the power of CodeLlama to address two critical use cases: Text-to-SQL generation, and bug fixing and vulnerability remediation.
Participants will need to create a HuggingFace account for the hands-on session.
Dr. Asankhaya Sharma is the Co-Founder and CTO at Patched.codes, a deep-tech startup with the vision to automatically fix every software vulnerability in code. Prior to Patched.codes, he was the Global Director of Innovation and Incubation at Veracode and the Head of R&D at SourceClear. He has a PhD in Computer Science from NUS.
2) Hacking My Way Through NUS: Prototyping to Real-World Usage
Discover Raivat’s experiences (from NUS class attendance bots to COVID food delivery systems), behind-the-scenes stories, and transitioning from prototypes to real-world applications, even earning academic credit for ‘failed’ hacks.
Raivat is a Founding Engineer building Large Language Models (LLM) in the education space.
3) DuckDB: A new kind of analytical database system
Learn about DuckDB, an in-process analytical database management system. DuckDB is open-source and highly portable with integrations for Python, R, Java, Julia, and 10+ other languages. This talk will introduce the DuckDB system, explain its key design decisions. Gábor will demonstrate how DuckDB works and showcase its innovative extensions to the SQL language.
Gábor Szárnyas is Developer Relations Advocate at DuckDB Labs. He obtained his PhD in software engineering in 2019 and spent 3 years as a post-doctoral researcher at CWI in Amsterdam, working on graph data management techniques.
👋 See you there!