Vishesh Paliwal

• Optimized Apache Druid segment layout and ingestion to sustain 5B+ logs/week, absorbing traffic spikes from 20M to 300M logs/hour (a 1400% surge) without query degradation.
• Working on request-response evolution — how API request and response shapes drift across services over time, in Kong's observability data plane.
• Working across the observability stack in Go: Kubernetes (AWS EKS), Kafka, Zookeeper, Druid, and Datadog.

• Avira, a RAG chatbot for bioprocess scientists. Query decomposition, Gemini File Search for retrieval, page-accurate citations, three retrieval modes (strict, augmented, rewrite).
• A multi-agent debate system on AgentScope: 7 specialist agents, moderator-driven consensus, persistent cross-session memory via ReMeLight + ReMe.
• Cloud Run deployments, PostHog AI for tracing.

• CI/CD on CircleCI + a GitHub Probot app. Deploy lead time down 25%.
• Fullstack observability with OpenTelemetry and Kloudmate. MTTR down 30%.
• HTML export service for motion ads, with a recovery-based DP compression pass. Output stays under 150KB without visible quality loss.
• PSD and Figma import plugins. Layer-by-layer parsing of images, text, and shapes into Rocketium's schema.

• Mentored 52 students through Data Structures & Algorithms.

TypeScript

• A small RAG pipeline I built while learning — chunking, embedding, retrieval, generation.

TypeScript

• An n8n community node that exposes memu's memory primitives to automation workflows.

Node.js · Express · MongoDB

• Backend for a video-sharing platform. Auth, uploads, comments, subscriptions.

Kotlin Multiplatform · Compose

• An AI jury simulator for Android and Desktop. Multi-agent personas, streaming responses, persistent debate history. A side project.

CGR 9.14

CGPA 7.6