Building a Text-to-Speech System for 25K Users (Without Melting a Server)
11 min read
You make your own luck if you stay at it long enough.
I build from zero. Currently a Member of Technical Staff at Mem0, working on the universal memory layer for AI agents. Whether it's frontend, backend, full-stack applications, or AI-powered experiences, I work across the entire development lifecycle. From UI/UX to deployment to user feedback, I care less about technology debates and more about delivering results that people love using.
Professional Experience
Mem0
Member of Technical Staff
Apr 2026 – Present
Mem0 (Intern)
Software Engineering Intern
Mar 2026 – Apr 2026
- •Authored 43 PRs (95% merge rate), 39 commits, 15K+ lines across 169 files — Python SDK, TypeScript SDK, docs, and CI.
- •Triaged issue backlog from 400+ to under 100, same for open PRs. Categorized, closed stale items, linked duplicates, routed actionable work to right contributors.
- •Fixed a critical data-loss bug where delete_all(user_id=...) destroyed the entire vector collection instead of filtered records — affecting all self-hosted users.
- •Built TypeScript SDK test infrastructure from scratch — 474 unit tests across 26 suites + 29 live integration tests against real Mem0 API, CI pipelines for Node 20 & 22.
- •Resolved 18 bugs across the stack: MongoDB metadata loss, Ollama URL misconfig, Gemini safety-filter crashes, LLMReranker silent fallback, tsup peer-dep bundling.
- •Reviewed 47 PRs from community contributors with detailed line-level feedback — guided MiniMax LLM, Apache AGE graph store, and REST server auth to merge-ready quality.
- •Drove community engagement — 34 external contributors, 57 PRs in 3 weeks, 72 issues resolved repo-wide.
- •Expanded docs with 3 OSS cookbook tabs, Vibecoding guide, MiroFish swarm-memory cookbook, and a full skills for mem0 (3,261 lines).
- •Patched CVE-2026-0994 (protobuf CVSS 8.2 HIGH), updated deprecated Google embedding models, migrated Ollama to current API endpoints.
- •Overhauled GitHub issue templates, configured stale-bot automation, added auto-labeler for 9 component types.
Turbo ML
Software Engineering Intern (AI)
April 2025 – July 2025
- •Implemented a reminder system using Redis Sorted Sets and a custom daemon to handle time-based task execution under 24-hour constraints.
- •Integrated multiple APIs including Swiggy for food delivery and dine-out services, and Blinkit for grocery delivery, while retaining Google for general recommendations and providing user choice flexibility.
- •Designed and implemented support for WhatsApp Business API commands (/help, /reset, /new), including runtime-configurable bot initialization based on environment variables.
- •Added Azure Blob Storage integration for file handling and message reaction management in WhatsApp bot.
- •Ensured WhatsApp message compliance by chunking outputs exceeding 4096 characters.
Sponsors
Buy me a coffee
Coding Stella
Suhail Roushan
Juan Daniel Martínez
GitHub Contributions ● @KartikLabhshetwar
Stack I use
Technologies I work with to build products that solve real problems
Open Source Contributions ● April 26, 2026
If you've read this far, you might be interested in what I do.
