About me

Software Engineer with 3+ years of experience designing and implementing intelligent systems, scalable web backends, and municipal/enterprise automation. Specialized in dotnet Core and Python, with a strong focus on RAG (Retrieval-Augmented Generation) architectures, AI Agents & MCP workflows, local ASR (Speech Recognition), and Persian NLP.

Technical Lead for 'Ravin' — the first specialized municipal and urban-planning AI assistant in Iran, deployed fully on-premise.

3+ Years of Experience
AI Lead Ravin AI Assistant
100% On-Premise Deployments

What I'm Doing

  • AI Systems & RAG Pipelines

    Designing deterministic RAG workflows, vector search, custom knowledge bases, and agentic integrations (MCP).

  • Scalable Backend Architecture

    Building resilient RESTful APIs, distributed microservices, and database schemas with dotnet Core, Python, SQL Server, and MongoDB.

  • Speech & Persian NLP (ASR)

    Deploying on-premise Sherpa/Gyro ASR engines and NLP models tailored for complex Persian administrative domains.

  • Identity & Access Management (IAM)

    Designing and implementing IDP & SSO systems with OAuth2, OpenID Connect, Keycloak, IdentityServer4, and OpenIddict.

Resume

View Resume Download

Education

  1. Isfahan University Of Technology

    2022 — Present

    Studying Computer Engineering

  2. Harati High school

    2019 — 2022

    Studied Mathematics in Harati high school

Experience

  1. Senior Backend & AI Engineer | Project Lead (Ravin)

    Nov 2023 — Present

    Nosaz Mohaseb Safahan — Municipal Automation Systems

    Architected and led the development of Ravin AI Assistant for municipal and urban planning workflows.

    Engineered zero-hallucination RAG pipelines with semantic search across extensive municipal regulations.

    Integrated local real-time ASR (Sherpa/Gyro) for Persian speech recognition in noisy office environments.

    Orchestrated fully on-premise (air-gapped) deployments ensuring data privacy and zero cloud dependency.

    Implemented observability and distributed tracing with OpenTelemetry and Grafana.

Tech Stack

AI / RAG & Agents

  • RAG
  • AI Agents
  • MCP
  • Semantic Search
  • Vector DBs
  • PyTorch
  • Persian NLP
  • ASR · Sherpa/Gyro

Backend & Distributed Systems

  • C#
  • dotnet Core
  • ASP.NET Core
  • Python
  • REST APIs
  • Microservices
  • Design Patterns

Data & Databases

  • MongoDB
  • SQL Server
  • Entity Framework
  • Data Modeling

DevOps & Observability

  • Docker
  • OpenTelemetry
  • Grafana
  • Linux
  • On-Premise Deployment

Security & Identity (IAM)

  • OAuth2
  • OpenID Connect
  • Keycloak
  • IdentityServer4
  • OpenIddict

Portfolio

  • Enterprise 100% Air-Gapped
    AI Systems / RAG

    Ravin — On-Premise Municipal AI Assistant

    The first specialized municipal AI assistant in Iran with Persian NLP, RAG, and Speech-to-Text, deployed entirely on-premise.

    The Challenge & Hard Constraints

    Municipal inquiries required interpreting complex, localized urban planning regulations with zero external cloud dependencies, no internet access (strict air-gapped security), and limited on-premise compute/VRAM.

    Engineered Pipeline

    Persian Voice/Text -> Offline ASR (Sherpa) -> Semantic Chunking & Vector Search -> dotnet Orchestrator -> Local LLM + Strict Guardrails

    Key Architectural Decisions & Evidence

    • Decoupled Execution dotnet Core handles deterministic business logic & API gateways; Python manages local model inference & embeddings.
    • Zero-Hallucination RAG Hybrid retrieval combining exact rule matching with dense vector embeddings stored in on-premise storage.
    • Resilient Audio Processing Low-latency streaming ASR tuned for noisy administrative environments.
    • C#
    • Python
    • RAG
    • ASR
    • MongoDB
  • Open Source
    Backend · Web Application

    Leave Management System

    Employee leave management web application built on dotnet Core with MongoDB persistence.

    • dotnet Core
    • C#
    • MongoDB
    View on GitHub
  • Backend · E-Commerce

    SibSalamat Online Pharmacy

    Online pharmacy platform built with dotnet 7 and MongoDB as the core persistence layer.

    • dotnet 7
    • MongoDB
    Repository private — available on request
  • Open Source
    Desktop · GUI

    IUT Messenger

    Desktop messenger application built with C++ and Qt 6 — IUT Advanced Programming final project.

    • C++
    • Qt 6
    View on GitHub