One MCP connection for actions across multiple apps
Llama 3.2 3B reached 79% at around 4.1GB VRAM
AI memory multitiering: Treat VRAM, RAM, and NVMe as placement tiers for the same model weights
No single model is responsible for everything. Qwen generates, Jev reviews, deterministic code controls execution, and Vercel eve manages the agent loop and approval flow.
I am a founder, CEO, and AI/ML community builder based in Bengaluru, India. My career spans developer advocacy, open‑source work, and growth marketing for AI‑native developer tools. I have built a personal audience of 200K+ developers across platforms and run the “AI Engineering” publication with 24,000+ readers.
Founder & CEO – Devable (Jul 2025 – Present). I build “growth engines” that turn technical content and channel strategy into reliable developer‑acquisition drivers. Clients include LandingAI, Orq.ai, Kilo Code, Glean, Superlinked, Minimax AI, and Mistral AI. Contact: sumanth@devable.ai
Author & Founder – AI Engineering (Apr 2025 – Present). The publication covers LLMs, AI Agents, and Retrieval‑Augmented Generation (RAG) and reaches 24,000+ AI/ML builders.
Machine Learning Developer Advocate – Clarifai (May 2023 – May 2026). I produced technical content, tutorials, demos, and a weekly developer newsletter, and represented the company at hackathons.
200K+ developers across platforms
In a market‑basket analysis project I improved model accuracy from 78 % to 82 % after hyper‑parameter tuning.