Software Development Engineer @ Innovativus  ·  B.Tech CSE @ VIT  ·  GSoC Contributor  ·  Full-Stack · AI/ML · Web3 · DevOps

     

 

🎙️ Voice AI	🤖 AI Agents	⚡ Throughput	🔗 On-Chain
STT → LLM → TTS	80%+ task success	300+ msg/sec	99.99% tx
cloning + Twilio calls	autonomous coding agent	real-time exchange	Solana DApp

👋 About Me Software Development Engineer at Innovativus Technologies — promoted from intern to full-time in 4 months. I build across four domains, going deep in each: production web apps, LLM-powered & voice AI systems, on-chain programs, and cloud-native infrastructure. Right now I'm building a production Voice AI platform: real-time two-way voice conversations with AI assistants, zero-shot voice cloning (F5-TTS + Supertonic), a Twilio phone assistant, and a RAG knowledge base — so users talk to their own data in any voice they choose. 🔭 Currently → SDE @ Innovativus · Voice AI platform 🎙️ Shipping → voice cloning · Twilio assistant · RAG 🌱 Learning → LLM internals · agentic systems 🤝 Open source → GSoC · p5.js · n8n 💬 Ask me about → LLMs · TTS · Next.js · Solana · K8s 📫 Reach me → yuvrajsingh9027249999@gmail.com ⚡ Fun fact → I clone voices & fine-tune LLMs for fun

⬆ my head, visualized — an LLM core with a voice waveform

🛠️ Tech Arsenal

💻 Languages

🤖 AI / ML · LLMs · Voice

 

🎨 Frontend	⚙️ Backend & Databases
☁️ DevOps & Cloud	⛓️ Web3 & Tooling

🎙️ Now Building — Voice AI Platform

Production voice AI @ Innovativus — talk to AI assistants grounded in your own knowledge base, in any voice you choose.

Capability	How it works
🗣️ Real-time voice conversations	Two-way voice chat with AI assistants over a live STT → LLM → TTS loop
🧬 Zero-shot voice cloning	F5-TTS + Supertonic neural TTS — natural cloned speech from seconds of reference audio
☎️ Twilio phone assistant	Automated inbound/outbound calls with configurable voice personas
📚 RAG knowledge base	Conversations grounded in the user's own documents & data

🚀 Selected Work

Project	Domain	Impact	Stack
🤖 Autonomous AI Coding Agent	AI/ML	80%+ task success · −40% hallucinations	Python · LangGraph · RAG · pgvector
🧠 LLM Fine-Tuning Pipeline	AI/ML	+25% accuracy · 15+ checkpoints	PyTorch · Hugging Face · LoRA · RLHF
⚡ Real-Time Crypto Exchange	Full-Stack	300+ msg/sec · 1,000+ clients	Next.js · WebSockets · Redis · Postgres
🎨 Doodle Space	Full-Stack	500+ concurrent · <100ms sync	Turborepo · WebSockets · Prisma
🔗 Solana DApp & Launchpad	Web3	99.99% tx success	React · Web3.js · SPL · Anchor
🏆 Contest Tracker	Full-Stack	CF · CC · LeetCode	Next.js · shadcn/ui · TypeScript

🌐 Open Source & Achievements

   

🏅	Highlight
🔀	Merged PRs into p5.js / Processing & n8n across 2 open-source orgs as a GSoC contributor
🤖	Shipped 4 production-grade AI projects (agents, RAG, fine-tuning, evals) in a 20+ module AI/ML program
🧱	Built 10+ full-stack apps in the 100xDevs Full-Stack, DevOps & Blockchain Cohort
🇮🇳	Proposed a citizen grievance platform at Smart India Hackathon 2025 (national level)

📊 GitHub Stats

 

🏙️ 3D Contribution City

_{my year of commits, rendered as an isometric 3D city — regenerated daily by GitHub Actions}

🐍 Contribution Graph

"Give AI a voice, give the web real-time speed — and make it all run at scale."