a16z Podcast | AMPLab, the Power of Open Source, and the Future of Systems Software

Amplab Origins and Model

• Founders Michael Franklin and Jan Stoica established Amplab after returning from two-year industry stints, seeking the high-velocity environment of startups rather than traditional academic pacing.
• The lab emerged from the "Rad Lab" initiative, merging expertise in systems and machine learning to address the emerging "big data" revolution.
• Funding is split 50/50 between government grants (NSF, DARPA) and corporate sponsors; the lab currently has 30 industry sponsors.
• All research is released as open-source software, with no IP rights granted to sponsoring companies.
• Corporate partners engage early in the research process to provide feedback on ideas and supply real-world problem statements.
• The university avoids patenting software, arguing that open-source brand building generates more significant industrial revenue than patent licensing.

Success Metrics and Outcomes

• The lab produces a high volume of groundbreaking startups, including Mesos, Apache Spark, and Tachyon (now Tachyon Nexus).
• Recent academic recognition includes two of the three global ACM Dissertation Awards going to Amplab researchers: Matej Zaharia (Spark) and John Duchi.
• The model relies on "traction" and "passion" as primary indicators for idea viability, rather than pre-defined commercialization plans.
• A central feature of the environment is the physical integration of faculty and students in shared spaces, removing traditional office barriers.

Investment Philosophy (A16Z/Peter Levine)

• A16Z views Amplab as a centralized mechanism to evaluate a full stack of emerging infrastructure projects simultaneously.
• Investment criteria prioritize companies founded by the original inventors of the open-source code, rather than those that are "forks" of the project.
• Investors believe the "flat memory" future—where RAM replaces hierarchical storage—will drive the next generation of database and storage architectures.
• A key investment heuristic involves identifying what will fill the gap if current dominant platforms (e.g., cloud, VMware) cease to exist.
• Successful academic entrepreneurs are defined by deep domain expertise combined with the "coachability" to learn business functions like hiring, sales, and marketing.

Hao-Yuan Li and Tachyon Nexus

• Li founded Tachyon Nexus based on Tachyon, a memory-centric distributed storage system designed to support the future of in-memory computing.
• Tachyon was developed to fill a missing layer in the Amplab ecosystem, complementing existing projects like Apache Spark and Apache Mesos.
• The transition from researcher to founder involved a significant shift from individual coding to team building, hiring, and process management.
• A primary current hurdle for the company is defining a commercial business model for the open-source project to avoid community resistance to future pricing.
• Li notes that while technical skills are taught in school, production-level system building and business strategy are largely learned through the entrepreneurial process.

Future Technology Trends

• The focus of big data is shifting from collection and batch processing to real-time analytics, machine learning, and deep learning.
• "Data science" has become a cross-disciplinary buzzword, attracting experts from fields like astronomy and not just computer science.
• Computing is moving from centralized cloud models toward "distributed computing at the endpoint," involving drones, IoT devices, and edge devices.
• Human computation is being re-integrated into automated systems to handle tasks where machine learning lacks sufficient accuracy.
• The next generation of the "Internet of Things" involves machines actively interacting with the physical world and the humans coexisting with them.

Strategic Insights on Academic Entrepreneurship

• Academic success in startups is often unplanned and organic, driven by faculty collaboration on shared interests rather than top-down mandates.
• The friction between student idea and user adoption has been drastically reduced by open-source platforms like GitHub and Amplab repositories.
• The "false dichotomy" between doing pure research and creating useful technology is rejected; impactful research naturally leads to commercial utility.
• A major challenge for PhD students transitioning to founders is the non-intuitive shift from "doing everything" as an individual to managing through others.
• Long-term success requires a willingness to prioritize hiring and coaching over personal technical output to avoid becoming a business bottleneck.