Nikhil Mungel
Head of AI R&D at Cribl
Nikhil is based in San Francisco, building distributed systems and AI teams at SaaS companies for over 15 years. His background spans AI, observability platforms, developer ecosystems, and high-scale consumer social products, with leadership roles at Substack, Splunk, ThoughtWorks, and most recently, Cribl where he currently serves as the Head of AI R&D.
Media
Expert commentary on how to secure and harden agentic software
Featured commentary in InfoWorld on AI skills for DevOps success in 2026.
Featured in StarCIO's 2026 predictions roundup alongside industry leaders.
Bylined article in Unite.AI on the significance of MCP for enterprise IT.
Quoted expert commentary in CIO.com on the shift to managing virtual agents.
Featured commentary in Centizen blog.
Guest appearance on 'Cribl: The Stream Life' Podcast (Episode 105) hosted by Bradley Chambers
Contributor to Chapter 10: 'Cribl, AI, and You' in this published book.
An introduction to Cribl Copilot hosted by Ed Bailey
Speaking
Invited speaker at RubyConf 2012.
Invited speaker at Splunk’s user conference, presenting how a commercial observability cloud platform is operated at scale
Invited speaker for Splunk's user conference, discussing how to analyze logs from microservices
Workshop at CodeDay Labs mentoring early-career engineers.
Featured speaker at CriblCon 2025, discussing schema transformations and human-in-the-loop AI.
BrightTALK Webinar co-presentation on onboarding data and generating queries with Copilot.
Invited speaker at AI Camp 2026 in San Francisco, discussing how to use AI to disrupt SaaS.
Research & Inventions
My work in the field of Distributed Systems, Observability, and AI has been cited in 845 publications by researchers and practitioners across the industry, including Palantir, Intel, Raytheon, SAP, IBM, Salesforce, Uber, Thoughtspot, Microsoft, Elasticsearch, Dell, Visa, JPMorgan Chase, Bank of America, Verizon, Snowflake, Oracle, Twitter, Amazon, and Cloudflare.
A gateway service for Distributed Systems that routes incoming HTTP requests to the right microservice based on the request method and headers, then creates and sends the corresponding outgoing request so the microservice can perform the requested action without being directly exposed. Developed before microservices were widely adopted, it helped shape how the industry routes and secures requests at scale in a microservices-based architecture.
A system that suggests structured data-retrieval queries (like SQL) as you type by recognizing key fields in your query, matching them to saved query patterns, using AI to recommend missing values, and showing the completed query for you to use.
A system that analyzes a partial structured data query (like SQL) as you type, runs related helper queries to learn about the available data, and then uses AI to recommend specific parameters you can add to complete and refine your query.
A system that suggests structured query parameters as you type by recognizing your partial query and using your organization’s past query usage to recommend relevant commands, functions, or field values to add.
A browser-based interface that uses AI to recommend and run multiple structured data queries side by side, showing each query and its results in separate areas and enabling you to execute one query based on the results of another.
Current Writing
Dives into what is team engagement and explores ways to track it.
Exploring how to assess engagement across teams of teams.
A look at the factors that impact team engagement.
Deeper dive into the factors that impact team engagement.
Guidance on conducting intentional one-on-one meetings.
The importance of lateral one-on-one meetings with peers and adjacent organizations.
In this post I share a practical way to judge whether an engineering org is successful by looking at the balance between outcomes and engagement, then unpack how vision, value-creating results, and engagement signals fit together.
Explore the central role of attention in our conscious thought and its impact on mental processes.
I conceived Cribl Copilot and guided my team to build it, and in this post I share how we made it explainable and steerable so humans stay in control.
In this post I argue that getting enterprise AI into production is an architecture problem, then share a practical decision guide for when prompting is enough, when fine-tuning earns its keep, and when training from scratch is worth the overhead.
Co-authored with Nick Heudecker. Analyzes the Model Context Protocol (MCP) as an existential threat and opportunity for SaaS vendors.
Technical tutorial on collecting real-time GPU telemetry for AI/ML workloads.
Technical Writing
Front page of Hacker News at #2 in May 2013.
Widely used technical resource with thousands of page hits per month.
Hobby Projects
The best tool to understand and learn about HTTP headers, used for debugging and optimizing HTTP requests and responses.