airhacks.fm podcast with adam bien

Adam Bien
airhacks.fm podcast with adam bien

Java, Serverless, Clouds, Architecture and Web conversations with Adam Bien

  1. 15 HR AGO

    TornadoVM: The Need for GPU Speed

    An airhacks.fm conversation with Michalis Papadimitriou (@mikepapadim) about: starting with Java 8, first computer experiences with Pentium 2, doom 2 and Microsoft Paint, university introduction to Object-oriented programming using Objects First and bluej IDE, Monte Carlo simulations for financial portfolio optimization in Java, porting Java applications to OpenCL for GPU acceleration achieving 20x speedup, working at Huawei on GPU hardware, writing unit tests as introduction to TornadoVM, working on FPGA integration and Graal compiler optimizations, experience at OctoAI startup doing AI compiler optimizations for TensorFlow and PyTorch models, understanding model formats evolution from ONNX to GGUF, standardization of LLM inference through Llama models, implementing GPU-accelerated Llama 3 inference in pure Java using TornadoVM, achieving 3-6x speedup over CPU implementations, supporting multiple models including Mistral and working on qwen 3 and deepseek, differences between models mainly in normalization layers, GGUF becoming quasi-standard for LLM model distribution, TornadoVM's Consume and Persist API for optimizing GPU data transfers, challenges with OpenCL deprecation on macOS and plans for Metal backend, importance of developer experience and avoiding python dependencies for Java projects, runtime and compiler optimizations for GPU inference, kernel fusion techniques, upcoming integration with langchain4j, potential of Java ecosystem with Graal VM and Project Panama FFM for high-performance inference, advantages of Java's multi-threading capabilities for inference workloads Michalis Papadimitriou on twitter: @mikepapadim

    1 hr
  2. 3 DAYS AGO

    Building AI-Native Code Platform With Java for Java

    An airhacks.fm conversation with Jonathan Ellis (@spyced) about: brokk AI tool for code generation named after Norse god of the forge, AI as complement to experienced programmers' skillsets, age and productivity in programming, transition from JVector to working on Cassandra codebase, challenges with AI in large codebases with extensive context, building tools for historical Java codebases, comparison of productivity between younger and older programmers, brute force coding vs experienced approach, reading code quickly as a senior skill, AI generating nested if-else statements vs better structures, context sculpting in Brokk, open source nature of Brokk, no black boxes philosophy, surfacing AI context to users, automatic context pulling with manual override options, importing dependencies and decompiling JARs for context, syntax tree based summarization, Maven and Gradle dependency handling, unique Java-specific features, multiple AI model support simultaneously, Claude vs Gemini Pro performance differences, Git history as context source, capturing commits and diffs for regression analysis, migration analysis between commits, AI code review and technical debt cleanup, style for code style guidelines, using modern Java features like var and Streams, Error Prone and NullAway integration for code quality, comparison with Cursor's primitive features, branching conversation history, 80% time in Brokk vs 20% in IntelliJ workflow, sketching package structures for AI guidance, data structures guiding algorithms, Git browser by file and commit, unified diff as context, reflection moving away from due to tooling opacity, Jackson serialization refactoring with DTOs, enterprise features like session sync and sharing, unified API key management, rate limit advantages, parallel file processing with upgrade agent, LiteLLM integration for custom models, pricing model based on credits not requests, $20/month subscription with credits, free tier models like Grok 3 Mini and DeepSeek V3, architect mode for autonomous code generation, code button for smaller problems with compile-test loop, ask button for planning complex implementations, senior vs junior programmer AI effectiveness, self-editing capability achieved early in development, no vector search usage despite JVector background Jonathan Ellis on twitter: @spyced

    1h 2m
  3. 17 JUN

    From Punch Cards (and Tapes) to Java

    An airhacks.fm conversation with Maurice Naftalin (@mauricenaftalin) about: Shelton Signet CP/M machine costing £3000 in the 1980s, discussion about the CP/M operating system which started in 1972, Maurice's early career teaching programming at Wolverhampton Polytechnic (now University), teaching Pascal programming language, creating a membership system for a political campaign using his first computer, Maurice's background as a chemist studying nuclear magnetic resonance (which later became MRI), learning fortran to process data using Fast Fourier Transforms, discussion about the NAG Library and challenges with array indices between C and Fortran, programming in the early days using punch cards and waiting hours for compilation results, the evolution from punch cards to paper tape which was more fragile, the role of punch operators who would type programs onto cards, Maurice's experience programming in assembler after learning Fortran, working at British Steel on an eccentric project to create a new programming language, moving to ICL (International Computers Limited) to work on the VMEB operating system with 15-16 protection rings, using traffic lights mounted on walls to indicate system status (red for down, amber for booting, green for operational), Maurice's interest in formal methods and the Vienna Development Method (VDM), working at Sterling University on formal specification and stepwise refinement, programming in HyperTalk for HyperCard in the 1990s, the Post Office Horizon scandal where a flawed computer system led to false fraud accusations against hundreds of sub-postmasters, Maurice's early Java programming creating a local information service distributed on CDs in the mid-1990s, discussion about offline-first principles and caching data that are still relevant today, Maurice being a "singleton" as the only Maurice Naftalin on the internet Maurice Naftalin on twitter: @mauricenaftalin

    1h 6m
  4. 18 MAY

    Accelerating LLMs with TornadoVM: From GPU Kernels to Model Inference

    An airhacks.fm conversation with Juan Fumero (@snatverk) about: tornadovm as a Java parallel framework for accelerating data parallelization on GPUs and other hardware, first GPU experiences with ELSA Winner and Voodoo cards, explanation of TornadoVM as a plugin to existing JDKs that uses Graal as a library, TornadoVM's programming model with @parallel and @reduce annotations for parallelizable code, introduction of kernel API for lower-level GPU programming, TornadoVM's ability to dynamically reconfigure and select the best hardware for workloads, implementation of LLM inference acceleration with TornadoVM, challenges in accelerating Llama models on GPUs, introduction of tensor types in TornadoVM to support FP8 and FP16 operations, shared buffer capabilities for GPU memory management, comparison of Java Vector API performance versus GPU acceleration, discussion of model quantization as a potential use case for TornadoVM, exploration of Deep Java Library (DJL) and its ND array implementation, potential standardization of tensor types in Java, integration possibilities with Project Babylon and its Code Reflection capabilities, TornadoVM's execution plans and task graphs for defining accelerated workloads, ability to run on multiple GPUs with different backends simultaneously, potential enterprise applications for LLMs in Java including model distillation for domain-specific models, discussion of Foreign Function & Memory API integration in TornadoVM, performance comparison between different GPU backends like OpenCL and CUDA, collaboration with Intel Level Zero oneAPI and integrated graphics support, future plans for RISC-V support in TornadoVM Juan Fumero on twitter: @snatverk

    1h 11m

About

Java, Serverless, Clouds, Architecture and Web conversations with Adam Bien

You Might Also Like

To listen to explicit episodes, sign in.

Stay up to date with this show

Sign in or sign up to follow shows, save episodes and get the latest updates.

Select a country or region

Africa, Middle East, and India

Asia Pacific

Europe

Latin America and the Caribbean

The United States and Canada