Aampe
An agentic intelligence layer for customer engagement. Each end-customer gets a dedicated AI agent that learns continuously from their behavior, what to say, when, how, and through which channel, compounding intelligence over time instead of resetting between campaigns. Deployed at 150M+ user scale, delivering 14-43% GMV lift and 57% higher notification opens across clients like Swiggy, Taxfix, Traveloka, Grab, and J.Crew. I designed the self-serve layer that lets marketing teams set strategy and guardrails without needing engineering support.
Case Studies (5)
Under NDA
These projects contain proprietary work. Request a password or enter one you've received.
Incorrect password. Please try again.
Surfacing reinforcement learning configuration to non-technical users
Aampe's multi-armed bandit system optimized millions of sends per week, but every agent evaluated message performance against the same reward signal. I designed a self-serve configuration layer that lets marketers define custom reward functions, mapping business-specific success criteria onto the underlying RL optimization without exposing model internals.
Detailed walkthrough available on call
Designing human-in-the-loop controls for autonomous agents
Autonomous agents had no mechanism to weight campaign importance. A routine newsletter and a high-stakes product launch received identical treatment in the optimization pipeline. I designed the priority system: five discrete tiers from Suppress to Maximize that feed into the agent's decision logic, giving marketers a direct lever to shape agent behavior without overriding the model.
Detailed walkthrough available on call
Visual query builder for complex user segmentation
Audience targeting was locked inside raw eligibility logic that only engineers could construct and debug. I researched segmentation patterns across the industry, then designed the Audience Builder from scratch: a visual query interface supporting multi-condition filtering, nested boolean logic, and real-time eligibility evaluation against Aampe's event-driven data model.
Detailed walkthrough available on call
Role-based access control for a multi-tenant AI platform
As Aampe's customer base scaled from startups to enterprise, the platform needed granular access control across every product surface. I designed the RBAC system: Viewer, Manager, and Admin roles with permission gating scoped to features, data visibility, and agent configuration, supporting multi-team environments within a single tenant.
Detailed walkthrough available on call
Ground-up rebuild of Aampe's core marketing interface
Aampe's largest platform initiative of the year: a ground-up rebuild of the Composer, the core interface where marketers configure agents, messages, and campaign logic. I maintained parallel Figma environments for exploration and production, introduced Figma Make for interactive hosted prototypes that replaced static handoffs, and led the full design token migration from the legacy Composer (CX) to the new system, remapping the color architecture across roughly 70% of the UI in a single sprint.
Bridging Figma components with engineering's Storybook primitives
Co-owned Aampe's internal design system, bridging Figma components with Storybook primitives in the engineering codebase. Established the component lifecycle process: local components are audited at sprint end, promoted to drafts, reviewed against Storybook parity, and classified as primitives or contextual patterns. This separation gave engineering a clear boundary between what lives in the shared library and what stays sprint-scoped.
What Else
Beyond the case studies above, I owned or drove initiatives across infrastructure tooling and design operations during my time at Aampe.
Bridging design tooling and production infrastructure with a Figma plugin
Aampe's message surfaces were designed in Figma but managed in the product, with no data connection between the two environments. I designed and shipped a Figma plugin across three phases that lets teams create, preview, and sync surfaces directly from the design tool into Aampe's messaging infrastructure, eliminating the handoff gap between design and deployment.
Designing Aampe's full website on Framer
Designed multiple pages for Aampe's full website redesign built on Framer: Solutions, Integrations, Blog (index, detail pages, and events), About (team section, careers), and navigation structure for both desktop and mobile. Worked through hero section explorations, layout systems, thumbnail aspect ratios, jumplink placements, and interactive states in Figma before collaborating with a brand agency and a Framer developer to bring everything to staging and production. The site launched in April 2025 as part of a broader brand overhaul.
Migrating and authoring Aampe's customer-facing documentation
Migrated Aampe's entire customer-facing documentation from ReadMe to Mintlify, restructuring the information architecture in the process. Authored feature articles explaining Aampe's product capabilities: how agents work, how reward signals are configured, and how the platform's optimization pipeline translates to marketer-facing controls. Called out by the CS Lead for the quality and depth of contributions to the Knowledge Hub.
