February 2026 Summaries

Nearly 100 students participated in the "From Events to Outcomes" Datathon, hosted in collaboration with the University of North Carolina Charlotte, using the Zerve platform to analyze real application data over a weekend. The event, which exceeded expectations, featured 23 teams who worked through the night, driven by genuine enthusiasm for solving data problems outside of a classroom setting. The winning team, comprising Meghana Sathi, Anisha Nannapaneni, Akhi Chappidi, and Chandra Siddharth Geddam, focused their analysis on platform user retention and presented concrete ideas for addressing user drop-off, which impressed the product team enough to revisit their findings multiple times. The event highlighted the value of having fresh eyes on a platform, as students often discovered insights that experienced teams might overlook, and the organizers are eager to repeat this successful collaboration.

A Zerve environment simplifies the use of Apache Spark by wrapping the Python Spark Connect client, allowing users to connect to Databricks without dealing with Java or JAR configurations. Introduced in Spark 3.4, Spark Connect separates the client and server, enabling a Python client that bypasses the JVM and communicates via gRPC. The environment includes both the databricks-sql-connector for SQL Warehouses, which follows the DB-API 2.0 spec, and the Spark Connect client for comprehensive data transformations using the DataFrame API. Zerve facilitates team collaboration by providing shareable environments, ensuring consistent dependencies across users, and allowing for the addition of extra packages via environment cloning. Credentials are directly transmitted from the Python client to Databricks, ensuring data privacy, and eliminating the need for Zerve to store or access data.

Andrew Engel, a former data science professional in B2B SaaS, co-founded NYST.AI, a startup focused on detecting concussions using phone camera video through pose estimation and machine learning. After spending six years in data science roles and experiencing disillusionment with the large language model market, Engel decided to address a real-world problem by leveraging his expertise in machine learning. NYST.AI aims to diagnose concussions by analyzing movement data captured from cell phone cameras, presenting a significant market opportunity in areas such as emergency medicine, sports, and military applications. The company is currently in the early stages, raising an angel round, and forming partnerships with medical institutions to gather necessary data and refine their models. Engel's shift from employee to founder has reinvigorated his passion for his work, as the potential impact of the startup offers a meaningful departure from his previous SaaS experiences.

Zerve has been actively participating in global hackathons, providing their platform to participants who have utilized it to create deployable solutions across diverse domains. At the Zerve x Unstop Hackathon in India, participants developed projects like an autonomous bidding engine and a subscription revenue forecasting system, showcasing Zerve's capability to bridge the gap between data science experiments and production microservices. Similarly, during the AI Datathon at Techfest, IIT Bombay, teams used Zerve to manage complex data preprocessing tasks efficiently, with winners demonstrating significant insights in health insurance claim predictions. Longevity Hacks in Berlin and Paris saw interdisciplinary teams creating AI agents for personalized longevity protocols, highlighting Zerve's role in fostering collaboration among diverse professionals. The ongoing Zerve.AI 2026 Hackathon on HackerEarth challenges participants to explore user behavior on data platforms, with substantial prizes and opportunities for showcasing their work. Across these events, participants consistently praised Zerve for enabling the transition from experimentation to finished, deployable systems, emphasizing its utility in real-world applications.

Choosing a data analytics platform requires careful evaluation beyond flashy demos, focusing on five critical dimensions: AI features, collaboration, production deployment, reproducibility, and performance. AI capabilities should extend beyond generic functionalities to truly understand data schemas and team-specific practices. Effective collaboration means real-time, multi-user editing and robust version control rather than basic link-sharing features. Production deployment is crucial, as most analyses fail to reach production due to the complexity of this step, necessitating seamless transitions from development to scheduled, reliable jobs. Reproducibility is vital to ensure consistent results across different environments, with platforms needing to manage dependencies automatically. Performance affects user behavior, where slow platforms lead to inefficient workarounds and missed validation steps, so stress-testing with real data and workloads is essential. Successful platform evaluations should prioritize workflow integration over feature checklists, involve diverse team roles, and focus on productivity impact rather than just cost, ensuring the chosen platform aligns with the team's specific needs and goals.

Zerve's AI agent offers a structured and interactive development approach by presenting a complete plan before execution, explaining its reasoning as it writes code, and tracking progress in real time. Users are given the opportunity to approve, reject, or revise the plan before any code runs, allowing them to direct the agent's tasks effectively. The agent provides reasoning messages alongside each code block, ensuring that the logic aligns with user intent and allowing for instant verification. As execution progresses, live progress indicators keep users informed about which tasks are active and how many remain. Once the plan is approved, the agent automatically creates workflow blocks, such as data generation and model training, on the canvas. After execution, a comprehensive summary is generated, including model results, performance metrics, and session insights, providing documentation for future reference without needing to reconstruct the workflow.

The conversation between the narrator and Satadru Sengupta explores the challenges and innovations in workflow automation, particularly focusing on the inefficiencies faced by service providers who are overwhelmed by multiple software solutions that do not integrate well. Satadru's new venture, Nimbus, aims to streamline these processes by offering a comprehensive system that handles tasks from booking to payment, as opposed to relying on disparate tools. The discussion also delves into the concept of "agent-native" design, where software is structured to be directly usable by AI agents rather than humans, exemplified by Nimbus's ability to automatically generate websites optimized for AI interaction. Additionally, the conversation reflects on the failure of Satadru's previous company, Dobby, due to market shifts and stresses the importance of questioning business assumptions even when things are going well. The sales approach for small businesses, contrasting with enterprise sales, is highlighted, emphasizing the need to offer immediate value to capture attention. Nimbus, currently in a pre-product-market fit stage, is navigating these challenges as it seeks to expand its customer base in the home services market.

The Men's Final Four Data Analytics Challenge is a unique student competition involving over 700 participants from colleges across Indiana, who are tasked with predicting the 68 seeds for the 2026 March Madness tournament over three months. Unlike typical classroom projects, the challenge presents messy data and undefined questions, requiring students to defend their analyses to audiences interested in conclusions rather than code elegance, thus mimicking real-world analytics work. The event aims to bridge the gap between academic learning and practical analytics by emphasizing the importance of asking the right questions and understanding the limits of one’s analysis. Zerve, the official LLM-based coding agent, supports students by allowing them to focus on problem-solving rather than dealing with technical infrastructure. This approach encourages participants to engage deeply with the analytical process, fostering curiosity and critical thinking skills essential for their future careers.