Skip to main content
STSoheil Tehranipour

Government & enterprise AI infrastructure

National AI Platform

LLM/VLM technical leadership for a multi-tenant platform at Sharif University of Technology

Year
2024 – present
Organization
National AI Platform, Sharif University of Technology
Industry
Government & enterprise AI infrastructure
Role
LLM/VLM Technical Lead

Personal vs. team contribution

Personal contribution

Designed the platform's high-level architecture (model APIs, development studio, monitoring, versioning, security and access control) and leads development and deployment of Persian-focused LLMs.

Team & organization

Implementation and operations are the work of a cross-functional team (model engineering, data engineering, DevOps); the platform belongs to Sharif University of Technology.

The challenge

Organizations and government bodies need capable Persian language models, yet building and running that infrastructure separately in each organization is costly and fragmented.

The solution

A multi-tenant platform that serves models as a shared, governed service: a unified model API, a development studio, model monitoring and versioning, and per-tenant security and access control.

Architecture

  1. Enterprise & government tenants
  2. API gateway & access control
  3. Development studio
  4. LLM/VLM model serving
  5. Monitoring, versioning & security

Technology stack

Pending confirmation This section will be completed once the information is confirmed.

Impact

5T+

tokens processed per month

Platform / team achievement

Lessons learned

Pending confirmation This section will be completed once the information is confirmed.