By playing, you agree to uListen's Terms and Privacy Policy and YouTube's Terms of Service.

Snapchat SVP of Engineering on How Everyone Ships Code Without Breaking Quality | Saral Jain | E304

In this episode of The Product Podcast by Product School, Carlos González de Villaumbrosia sits down with Saral Jain, SVP of Engineering at Snapchat, the last independent social platform operating at global scale, with 956 million monthly…

Saral JainguestCarlos González de Villaumbrosiahost

Jul 15, 202632mWatch on YouTube

Summaries, chapters, transcripts, and FAQs on this page are made by uListen, not YouTube.

CHAPTERS

Jump to publisher timestamps

Chapter summaries

Prototypes beat mockups: building is the new bottleneck breaker
Saral Jain explains that the team increasingly settles debates by building prototypes rather than relying on static mockups. He frames rapid prototyping as a way to let the best idea win, especially as AI reduces execution friction.
- •The bottleneck isn’t mockups; it’s quickly building something tangible
- •Teams show prototypes and let the best idea win
- •AI is changing how fast ideas can move from concept to working software
- •Role boundaries matter less than what people spend time doing
Meet Saral Jain and Snap’s scale as context for engineering choices
The host introduces Saral’s role and scope, and Saral describes Snapchat’s scale and engagement. This sets up why quality, speed, and platform leverage matter when shipping to a massive global audience.
- •Saral leads most of engineering for Snapchat
- •Snapchat is nearly 15 years old and nearing a billion monthly active users
- •Engagement is frequent, with the app opened many times per day on average
- •The conversation will focus on how the organization ships quickly without sacrificing quality
How Snap is organized: functional leadership and tight product-engineering-design partnership
Saral outlines a functional org model where functional leaders report to the CEO rather than using a GM structure. He describes how product, design, and engineering operate closely, with design strongly influencing ideas and execution.
- •Functional leads report to the CEO rather than a general manager model
- •The CEO plays a strong product/design role, with heads of product and design reporting to him
- •Product, design, and engineering collaborate closely and share idea ownership
- •Design and product are under the same leader, working as a tight trio with engineering
Designers (and others) shipping production code as roles blur
Saral says he’s become comfortable with non-engineers shipping code because role boundaries are changing. He emphasizes that Snap’s long-term platform investments help reduce the risk of outages or regressions as more people contribute.
- •Designers, design engineers, and product managers can ship production code
- •The focus is shifting from job title to actual work and impact
- •Snap’s codebase and platforms are built to limit the “blast radius” of changes
- •Cross-functional contribution is treated as a practical reality, not a novelty
Startup squads: lean cross-functional teams for zero-to-one bets
Saral describes Snap’s “startup squads” as small, cross-functional teams that operate like startups inside a mature platform. These squads reduce rigid role distinctions and focus on launching new initiatives quickly.
- •Startup squads include engineers, designers, product managers, and data scientists
- •Teams treat efforts as zero-to-one initiatives or new revenue initiatives
- •Role distinctions are intentionally less rigid inside squads
- •Saral says Snap invested in multiple squads and saw strong results quickly
Guardrails for quality at speed: CodePal AI review plus human ownership
Saral explains that quality is critical at Snap’s scale, so platform guardrails are a top priority. He describes an AI code review agent, CodePal, and clarifies that humans still own and approve production code today.
- •Quality is a top focus because Snap ships to a massive global audience
- •CodePal performs the first pass of code review and quickly covers most code changes
- •The agent can reason across many repositories and identify patterns humans may miss
- •Humans still own and manually review code before shipping, with gradual evolution over time
Why spin off new companies: different operating models, shared foundations
Saral explains the business rationale for spinning off efforts like Specs and Dotmo. He contrasts the needs of a mature platform with those of a new computing platform or capital-intensive generative AI work, while noting shared underlying technology.
- •Specs is positioned as a new computing platform requiring different thinking than a mature social platform
- •Snap has invested in AR for a long time and sees convergence with stronger AI and hardware
- •Dotmo is framed as a generative AI video-focused startup building on years of AR and model investments
- •Foundational technology can be shared across Snapchat and spun-out entities
Competing with giants: creativity, speed, and AI reducing execution bottlenecks
Saral reflects on Snap’s long-standing need to compete with far larger companies. He attributes success to creativity, moving fast, deep understanding of the community, and sees AI as especially beneficial because it reduces execution constraints.
- •Snap has historically competed against companies with far greater resources
- •The company’s DNA emphasizes creativity, speed, and community understanding
- •Saral sees AI as removing execution bottlenecks so ideas can ship faster
- •He frames this moment as especially advantageous for a creative organization
AI after restructuring: three verticals—users, employees, advertisers—and the case for smaller teams
Saral addresses restructuring as a business reality and describes how AI contributes to confidence going forward. He organizes Snap’s AI efforts into three areas and argues that smaller teams tend to be more nimble and fast-moving.
- •AI is used for user-facing product features, internal work efficiency, and advertiser/partner value
- •Progress across all three areas supports confidence in continued speed
- •Saral believes smaller teams are more flexible and effective
- •AI is positioned as a multiplier across the organization, not a single tool
AI features for Snapchatters: My AI and generative AI lenses
Saral explains that Snap doesn’t label things as “AI features” so much as features that add value. He highlights My AI as a personality-driven assistant and describes generative AI lenses, including an open-prompt lens experience.
- •Snap frames features by user value rather than whether they are AI-powered
- •My AI is described as a chatbot/assistant with a distinct personality and use cases
- •Generative AI lenses build on Snapchat’s camera-first experience and creative play
- •An open-prompt lens lets users transform what they see with text prompts, and Saral cites broad engagement with Gen AI lenses
Defending the moat: platform, network effects, and communication with close friends
Saral argues that copying features is less important than the underlying platform and network effect. He ties Snap’s product DNA to fun, casual communication with close friends and family, and says features succeed because they reinforce that behavior.
- •The moat is framed as platform plus network effects, not any single feature
- •Snap has repeatedly pioneered features that later become standard elsewhere
- •People primarily use Snapchat to communicate with close friends and family
- •Product development ties back to fun, casual interaction rather than public validation dynamics
Who Snapchat is for and how segmentation happens in practice
Saral says Snapchat is for everyone, especially at nearly a billion monthly users, and emphasizes high engagement. He explains that Snap considers value propositions per feature and sometimes lets the community reveal unexpected audiences, using Topic Chats as an example.
- •Snapchat is described as serving everyone across geographies and demographics
- •High engagement is highlighted as a core strength of the platform
- •Segmentation starts with each feature’s value proposition rather than rigid upfront assumptions
- •Topic Chats is cited as a feature whose usage patterns (e.g., around major events) help inform iteration
AI for advertisers: ROI-first targeting, measurement, creatives, and AI Sponsored Snaps
Saral emphasizes that advertisers don’t want AI for its own sake; they want performance improvements. He outlines AI across the advertiser journey and gives an example of AI Sponsored Snaps that enable chat-based brand interactions.
- •Ads remain central to revenue, alongside a growing subscription product
- •AI is used to improve advertiser ROI through targeting, measurement, and creative generation
- •AI Sponsored Snaps bring advertisers into the chat feed with AI-driven interactions
- •The value proposition ties advertiser products to how the community uses Snapchat
Turning AI adoption into business impact: Jobs To Be Done as the operating system
Saral describes an early phase where AI tool usage was high but prototypes were frequently discarded. He explains how leadership used a Jobs To Be Done framework to align AI work with real roadmap items, enabling measurable business impact.
- •Adoption of AI tools was widespread, but impact was limited when work stayed at the prototype stage
- •Leadership mapped “jobs” across functions and asked where AI can accelerate or rethink them
- •Roadmaps are aligned to these jobs so AI efforts attach to real deliverables
- •This approach creates guardrails and supports change management through training and enablement
Agents at Snap: horizontal vs vertical agents, with Casper as an AI teammate
Saral explains Snap’s mix of foundational agents used broadly and specialized agents for specific workflows. He highlights Casper, an agent embedded in existing tools that can be invoked to build from conversation context, alongside CodePal as the review layer.
- •Snap uses both horizontal (foundational) and vertical (function-specific) agents
- •Examples of vertical agents include ones focused on experimentation analysis, data science, and cost inefficiencies
- •Casper is described as a company-wide AI teammate available in tools like Slack and Jira
- •Casper can generate code based on conversation context, with CodePal reviewing changes as part of the workflow
Engineers prototyping in code and planning in a fast-changing AI environment
Saral describes a shift toward engineers building prototypes directly, supported by internal design systems and AI tooling. He also explains why planning has moved from long annual plans to more flexible quarterly planning with frequent check-ins.
- •Snap values product-minded engineers who ask not just how, but why
- •Prototyping often happens directly in code, supported by an internal design system
- •Multiple AI tools are used to accelerate prototyping and iteration
- •Planning is broken down quarterly with high adaptability, including frequent check-ins as conditions change

Publisher timestamps

Get more out of YouTube videos.

Read along on your phone, search it beside the video, or ask your favorite AI app what was said.