IBM at Boston Tech Week 2026

Learn about efficient AI inference and put it into practice in a hands-on workshop focused on fast, high-performance deployments with vLLM. Together, we’ll deploy a dense LLM with vLLM, benchmark performance using GuideLLM, and explore techniques like quantization and speculative decoding to improve throughput, latency, and overall efficiency. We’ll also showcase LLM Compressor, an open source library for model quantization. As part of the workshop, we’ll deploy a quantized model together, then measure the impact on the AI metrics that matter most in production deployments. The session will conclude with an open community discussion and networking.

Who should attend: AI developers, ML engineers, platform engineers, data scientists, and anyone passionate about AI, data, and computer science who wants to get hands-on with modern, open source LLM deployment and optimization techniques.

What to bring: A fully charged laptop!

Register Here: https://partiful.com/e/2CWzL7efN3h5MtSTO9nw

This is the first in a series of four "Build-n-Brew" sessions to be held in the afternoon of each day during Boston Tech Week, hosted at the MIT-IBM Computing Research Lab at "Pi on Main" (314 Main Street) in the heart of Kendall Square. Co-sponsored by The Open Accelerator, RedHat, and IBM, each session starts at 4pm and will be a hands-on "build" experience to learn and use the latest in open source AI tech, to be followed by a reception for a little nosh, "brew", and networking. This first session will be on vLLM.

NOTE: Doors open at 3:30p and space is limited to the first 50 to arrive - If you don't make it in time, we'll leave you with a small consolation prize and an invitation to join us at a future event.

Docling is an open-source document processing framework developed by IBM Research that converts unstructured documents (like PDFs, DOCX, PPTX, and images) into clean, structured, machine-readable data. It is designed to bridge the gap between messy, real-world business documents and AI.

In this build session, we'll be walking through the features of Docling. By the end, you will be able to use Docling to:

• Convert complex documents into various formats suitable for AI
• Understand the various methods of Chunking and how to use them
• Create a transparent and verifiable multimodal RAG application
• Navigate a parsed document's section tree with an LLM, without chunking or embeddings
• Deploy Docling as a scalable REST API service
• Compare token costs for PDF Q&A with and without Docling preprocessing
• Integrate Docling as a tool for AI agents using the Model Context Protocol (MCP)

Bring a laptop. The session is open to engineers, AI builders, and PMs who are interested in leveraging the latest open source tools for software development.

Register here: http://partiful.com/e/K0u54muKGZTpEnUju7Jb

This is the second in a series of four "Build-n-Brew" sessions to be held in the afternoon of each day during Boston Tech Week, hosted at the MIT-IBM Computing Research Lab at "Pi on Main" (314 Main Street) in the heart of Kendall Square. Co-sponsored by The Open Accelerator, RedHat, and IBM, each session starts at 4pm and will be a hands-on "build" experience to learn and use the latest in open source AI tech, to be followed by a reception for a little nosh, "brew", and networking. This second session will be on Docling.

NOTE: Doors open at 3:30p and space is limited to the first 50 to arrive - If you don't make it in time, we'll leave you with a small consolation prize and an invitation to join us at a future event.

Most agents die in staging. Here's how to build ones that don't w/ IBM Bob and WXO

The future of software is going to be built with AI agents, but most agents never make it to production today. The ones that do often take months to harden, and even then, it's unclear whether they are delivering the value they were built for, or quietly burning tokens.

In this hands-on session, we'll bring together two halves of the agent lifecycle. We will use IBM Bob to build an agent from natural language. Then we will deploy it to watsonx Orchestrate, and take you through the full agent-ops cycle -- observability, evaluation and optimization -- to make your agent production ready.

Bring a laptop. The session is open to engineers, AI builders, and PMs who are interested in building agents that do more than demo well.

Register here: https://partiful.com/e/jixyrcrLUUKcaFq0HYSJ

This is the third in a series of four "Build-n-Brew" sessions to be held in the afternoon of each day during Boston Tech Week, hosted at the MIT-IBM Computing Research Lab at "Pi on Main" (314 Main Street) in the heart of Kendall Square. Co-sponsored by The Open Accelerator, RedHat, and IBM, each session starts at 4pm and will be a hands-on "build" experience to learn and use the latest in open source AI tech, to be followed by a reception for a little nosh, "brew", and networking. This third session will be on Bob.

NOTE: Doors open at 3:30p and space is limited to the first 50 to arrive - If you don't make it in time, we'll leave you with a small consolation prize and an invitation to join us at a future event.

Learn about efficient AI inference and put it into practice in a hands-on workshop focused on fast, high-performance deployments with vLLM. Together, we’ll deploy a dense LLM with vLLM, benchmark performance using GuideLLM, and explore techniques like quantization and speculative decoding to improve throughput, latency, and overall efficiency. We’ll also showcase LLM Compressor, an open source library for model quantization. As part of the workshop, we’ll deploy a quantized model together, then measure the impact on the AI metrics that matter most in production deployments. The session will conclude with an open community discussion and networking.

Who should attend: AI developers, ML engineers, platform engineers, data scientists, and anyone passionate about AI, data, and computer science who wants to get hands-on with modern, open source LLM deployment and optimization techniques.

What to bring: A fully charged laptop!

Register Here: https://partiful.com/e/2CWzL7efN3h5MtSTO9nw

This is the first in a series of four "Build-n-Brew" sessions to be held in the afternoon of each day during Boston Tech Week, hosted at the MIT-IBM Computing Research Lab at "Pi on Main" (314 Main Street) in the heart of Kendall Square. Co-sponsored by The Open Accelerator, RedHat, and IBM, each session starts at 4pm and will be a hands-on "build" experience to learn and use the latest in open source AI tech, to be followed by a reception for a little nosh, "brew", and networking. This first session will be on vLLM.

NOTE: Doors open at 3:30p and space is limited to the first 50 to arrive - If you don't make it in time, we'll leave you with a small consolation prize and an invitation to join us at a future event.