Head of Engineering for AI at The Browser Company, leading the evolution from Arc to Dia—the next generation of AI-native browsers. Previously spent 6 years as a senior engineer at Instagram/Facebook, bringing expertise in building products at scale. At the AI Engineering Code Summit, shared critical lessons about integrating AI into beloved products and building infrastructure for rapid iteration.
Pioneering AI-Native Browser Development
Samir Mody leads AI engineering at The Browser Company, where his team is reimagining web browsing with artificial intelligence at its core. Following Atlassian’s $610M acquisition of the company, Mody has been instrumental in transitioning from Arc—a browser beloved for its UX—to Dia, an AI-native browser designed for speed, security, and personalization.
Professional Background
Previous Experience: Senior Engineer at Instagram/Facebook (6 years), working on large-scale consumer products and infrastructure. Education from UCLA (2009-2013).
Current Focus: AI integration in browsers, agentic behavior design, prompt engineering at scale, security in AI-powered consumer products, and building developer tools for rapid AI iteration.
Key Philosophy: Tools, Craft, and Speed
Mody’s approach centers on three principles that transformed The Browser Company’s AI development velocity:
10x Iteration Speed Through Custom Tooling - Built a prompt editor directly into dev builds, moving all prompts into development tools. This enabled engineers to iterate with full context immediately available, multiplying ideation speed by 10x. Reference: GEPA paper on arXiv discusses systematic prompt optimization approaches.
Model Behavior as Craft - Treats AI behavior design with the same rigor as traditional software engineering: intentional behavior design, quantifiable measurement, model steering techniques, and continuous refinement. Shifting from functional AI (responds to queries) to agentic AI (proactively suggests actions with rationale).
Security as Product Property - Addresses the “lethal trifecta” of prompt injection, data exfiltration, and behavior manipulation through human-in-the-loop confirmation for sensitive tool calls and wrapping untrusted context in protective tags. Acknowledges “there are no guarantees” but designs safety into the product from day one.
Notable Work & Contributions
- Arc Browser: Enhanced a beloved browser with AI capabilities while maintaining user trust
- Dia Browser: Building an AI-native browser from lessons learned integrating AI into Arc
- Developer Infrastructure: Built eval collection systems, data pipelines, and automation for prompt optimization
- GEPA Implementation: Applied systematic “hill climbing” approaches to prompt engineering at scale
Summit Presentation Highlights
Session: “From Arc to Dia: Lessons Learned in Building AI Browser” (November 20, 2025, 2:45 PM)
Core Insights:
- Purpose-built tooling for AI features multiplies iteration speed (10x improvement)
- Treat model behavior as a craft requiring discipline around design, measurement, and steering
- Security against prompt injection must be designed into products, not bolted on
- Organizations recognizing AI as a technology shift must “embrace it with conviction”
Technical Contributions: Discussed moving beyond traditional development processes to support rapid AI iteration, implementing systematic approaches like GEPA for prompt refinement, and building infrastructure for eval collection and automated testing.
The Arc to Dia Evolution
The transition represents applying lessons from retrofitting AI into a beloved product to building an AI-native browser from scratch. Both browsers target speed, security, and personalization, but Dia treats AI as a core architectural principle rather than a feature addition.
Impact & Relevance
Mody’s work is especially relevant for organizations building AI-native products, deploying AI at consumer scale, navigating organizational transitions around AI adoption, and evolving engineering culture to treat AI behavior as craft.
Notable Quotes
- “10x the speed of ideating and iterating” (on integrated prompt editors)
- “We are in the early days of model behavior” (on nascent AI behavior design)
- “When you recognize that it tech shifts, you have to embrace it with conviction”
- “Treating model behavior as craft” (on disciplined AI feature development)