Headline

Building AI-Powered Backend Systems @Taiyari Karlo | Ex-Oracle MTS | Kafka • Kubernetes • LLM Pipelines

About

Backend engineer with 4+ years building distributed, high-availability systems — currently designing AI-powered content pipelines and live-streaming infrastructure at Taiyari Karlo. At Oracle, I containerized Kafka infrastructure (90% faster deployments), built multi-broker clusters with zero-message-loss failover, and shipped an AI-assisted vulnerability scanner that cut manual security review time by 95%. My profiling and performance-analysis tooling reduced IPC latency by 23% and cut regression-diagnosis time from hours to minutes. Now at Taiyari Karlo, I'm building end-to-end AI pipelines (Java, Spring Boot, async queues, DeepSeek, Qwen) that turn raw lecture video into structured study material — MCQs, summaries, and practice tests — with zero manual authoring. I also designed the backend for a live classroom streaming platform (RTMP ingest, HLS playback) supporting thousands of concurrent viewers via CDN fanout.

Location

Bikaner

AI stack

anthropic apis

claude code

claude

codex

langgraph

openai

Work experience

AI Software Developer at Taiyari Karlo, Jan 2026 - Present · 9 mos, Remote, • Built an end-to-end AI content-generation pipeline (Java, Spring Boot, async queues, DeepSeek, Qwen) that auto-generates MCQs, summaries, and practice tests from lecture videos, eliminating manual content authoring for teachers. • Designed and shipped the REST API layer serving AI-generated content to production, handling generation status, retrieval, and delivery to teacher/student clients; implemented chunk-based transcript generation to optimize AI processing throughput. • Designed the backend for a live classroom streaming platform using RTMP ingest and HLS playback, supporting thousands of concurrent viewers via CDN fanout.

Member of Technical Staff at Oracle, Jul 2022 - Oct 2025 · 3 yrs 4 mos, India, • Containerized Oracle Kafka Server (OKS) using NanoVOS base images and Kubernetes (Minikube & OKE), reducing deployment time by 90%. • Designed and deployed multi-broker Kafka clusters with StatefulSets, Headless Services, and shared storage, ensuring persistent identity, seamless scaling, and zero-message-loss failover for high availability • Developed Kafka TEQ Wrapper to enable seamless publish-subscribe messaging between Kafka client applications and Oracle Database via PL/SQL procedures, improving modularity and maintainability. • Reduced IPC latency benchmark by 23% and improved regression detection by developing profiling scripts to track function coverage, identify gaps, and automate performance analysis across builds. • Built an Automated Performance Analysis System that captures execution traces, generates analytics reports, and displays dashboards to quickly identify and track performance issues at the function level. • Reduced manual effort and improved debugging efficiency by enabling BDCell to leverage internal Oracle debugging tools by enhancing KGRK integration which included initializing Oracle subsystems (KGH, DBGT, GP), adding diagnostic support for native threads, and building unit-test utilities. • Implemented automated security scanning and fuzzing workflows to proactively detect vulnerabilities, uncovering critical security flaws (CVE-2021-3520, CVE-2022-2132) and eliminating manual security review efforts through complete automation.

Server Technology intern at Oracle, May 2021 - Jul 2021 · 3 mos, • Implemented publish-subscribe messaging using Oracle Cloud Infrastructure(OCI) Streaming Service via REST API’s, replacing the dependency on Confluent API’s • Developed and tested kafka client integrations (producers/consumers) with librdkafka and libcurl, enabling end-to-end stream operations and laying ground work for multi stream scalability.

Education

Bachelor of Technology, Computer Engineering, National institute of technology Calicut

Public links

github: https://github.com/tarunkansal1