Krishna Mallam.
About Krishna Mallam
CTO at Medialogic AI. For two years now I've been shipping agentic platforms to regulated European enterprises in Banking, Insurance and Telecom: open-source LLMs (Llama 3, Mistral, DeepSeek, Qwen3, Kimi K2.6) on EU-owned bare metal, enterprise RAG with grounded citations, human-in-the-loop governance. Hybrid infra on Hetzner and OVH with Proxmox, up to ~40% lower TCO than hyperscaler GPU instances at sustained scale. GDPR by design.
# positioning: Sovereign AI for regulated EU enterprises · EU jurisdiction, not just EU region · up to ~40% lower TCO than hyperscaler GPU instances at sustained scale
“A US cloud in Frankfurt is still a US cloud. The CLOUD Act and FISA 702 follow the provider, not the server, so a US parent can be compelled wherever your data sits. We remove the parent: open-weight LLMs on EU-owned bare metal, with weights, logs and keys under EU jurisdiction, not just EU residency.”
Currently working on
- AI-native engineering velocity. Claude Code and Cursor drive the Medialogic engineering loop. Role-specific CLAUDE.md / AGENTS.md files per repo (architect, backend, frontend, infra) so the same instructions guide any coding agent we run.
- Enterprise RAG, governed. Multi-tenant retrieval over PDFs, SQL, SharePoint and certified PEC mail. Grounded citations, tenant isolation. Banking, Insurance and Telecom in production.
- Sovereign hybrid bare-metal. OVH France and Hetzner with Proxmox, Ollama / vLLM on bare-metal GPUs. Data stays in EU, ~40% cheaper than hyperscaler GPU rent.
- Cross-region engineering. Engineering team across Rome and Hyderabad. One roadmap, shared CI, daily overlap between time zones.