Prakhar Gandhi
Lead AI Engineer @ Tiger Analytics | Google Developer Educator (Compose & KMP)
Bengaluru, India
Actions
Prakhar Gandhi is a Lead AI Engineer at Tiger Analytics and a Google Developer Educator specializing in Jetpack Compose and Kotlin Multiplatform. A prolific builder with deep roots in modern mobile architecture and production AI systems, he bridges cutting-edge machine learning with scalable, cross-platform client applications. His work spans intelligent systems, hardware integrations using NFC tech, and early-stage product engineering with Heyum. Outside the terminal, Prakhar hosts The Broadcast Bros podcast, dissecting technology, product development, and engineering culture. He regularly shares practical, real-world patterns to help developers build fast, resilient, and intelligent cross-platform apps.
Area of Expertise
Topics
The Jailbreak You Didn't Know You Taught It: Adversarial Instruction Under Academic Framing
Safety-aligned LLMs refuse harmful requests outright — but reframe the same request as academic, pedagogical, or research-oriented, and the model often complies in full detail. This talk walks through a documented vulnerability class (10 reproducible, controlled experiments, now confirmed reproducible on a live production model) where "teaching" framing bypasses refusal behavior that direct requests trigger. I'll show the experiment design, the failure pattern across models, why current alignment techniques miss this class of attack, and what builders of LLM-powered products need to check for before shipping. No slideware theory — this is original, peer-citable research (published on SSRN) distilled into what you can act on Monday morning.
Beyond the Hello World: What Passkey Tutorials Don’t Tell You
Every passkey tutorial stops at the happy path: one device, one browser, sign-in works. We went past that on purpose. We built a passkey flow across Android and the web, then tried the things real users do: cancel the sheet, delete a key on one device, swap phones, reject the biometric prompt. This talk is what broke, why the docs did not warn us, and the patterns we landed on for fallback, recovery and cross-platform state.
The slides mostly survive as they are. Change “field notes” to “what we hit”, and “the recovery nightmare” stays because it is a design argument, not a claim of experience.
Server-Driven Android: How Much Control Is Too Much?
The power to instantly alter app behavior or layout without waiting for a Google Play Store release is incredibly tempting, but handing total control to the backend frequently introduces severe security vulnerabilities, data parsing bugs, and backward-compatibility chaos. When a server payload changes unexpectedly, older client versions still live in production can instantly crash or misbehave.
Instead of showing generic framework setups or code walkthroughs, this circle discussion focuses on the operational realities of server-driven architectures. We will discuss concrete experiences regarding what happens when remote configurations corrupt client states, how to handle edge cases where the backend tries to trigger actions the app cannot support, and where developers draw a firm boundary between server-driven UI and server-driven logic.
NFC in Modern Android: Building Cross-Platform Experiences with Jetpack Compose & KMP
NFC can turn an ordinary Android interaction into a simple “tap and go” experience—but building NFC-powered applications that are maintainable, testable, and ready for modern Android architecture is a different challenge.
In this session, we will build an NFC-driven Android experience using Jetpack Compose and Kotlin Multiplatform (KMP), exploring how platform-specific NFC capabilities can be cleanly separated from shared application logic.
Attendees will learn how to move beyond basic NFC tag reading and design an architecture that can support real-world NFC products and experiences.
We will cover:
1. How Android's NFC stack works and how to integrate it cleanly with a Compose application
2. Designing an NFC abstraction that keeps platform-specific code isolated using KMP
3. Managing NFC state, lifecycle, permissions, and asynchronous events in a Compose-first architecture
4. Designing reliable NFC flows for reading, writing, validation, and error handling
5. Testing and structuring NFC functionality so that the application remains maintainable as it grows
The goal is to leave attendees with a practical architecture they can adapt to their own NFC-enabled Android and KMP projects.
Mobile + AI in 2026: Build Features, Not Wrappers
There are 10,000 ChatGPT wrapper apps. None will matter in a year. Here's what will.
The AI mobile apps that win aren't just API integrations—they're using AI as a substrate to enable entirely new behaviors impossible without it. This session breaks down the difference.
We'll analyze:
1. Wrappers vs Features: Why "ChatGPT in my app" is already commoditized, but "AI that understands my users' context" is defensible
2. The Data Moat: How mobile apps can collect unique training data (sensors, usage patterns, behavioral signals) that cloud AI can't access
3. Agentic Mobile: Moving beyond chat—AI that takes actions, automates workflows, and operates autonomously within your app
4. The Multimodal Advantage: Mobile has camera, mic, location, motion—use them. Text-only AI is leaving money on the table.
5. Case Study Breakdown: Apps doing this right (and what makes them hard to copy)
For builders who want to understand where mobile AI is actually going, not just where the hype is today.
droidCon India 2026 Sessionize Event Upcoming
Please note that Sessionize is not responsible for the accuracy or validity of the data provided by speakers. If you suspect this profile to be fake or spam, please let us know.
Jump to top