Mohadese Zamani GitHub

AI Engineer building stateful LLM workflows and applied ML systems.

I turn natural-language analysis requests into typed plans, validated artifacts, and user-facing interpretations.

See selected work GitHub
  1. Analyze this workbook and explain the drivers.
  2. clarify_intent() build_plan() route_review()
  3. contract: AnalysisPlan schema: valid
  4. Reviewed by: Analyst Decision: approved
  5. Validated artifacts ready for interpretation.
[ state saved ]

Selected work

Systems and architecture writing that show how I structure AI workflows around explicit state, contracts, validation, and review.

[ THE WORKFLOW NOTEBOOK ]

AI-Powered Multi-Agent Data Analysis System

A stateful workflow that turns natural-language analysis requests into clarified intent, execution plans, validated artifacts, and user-facing interpretations.

Python · LangGraph · FastAPI · Pydantic · JSON Schema

View case study

Financial Assistant: Spreadsheet Intelligence and Profiling Service

A processing pipeline for complex financial workbooks covering extraction, region reconciliation, normalization, profiling, classification, and semantic enrichment.

Python · pandas · PyArrow · RabbitMQ · PostgreSQL · MinIO · Docker

View repository

Designing an Auditable Hybrid Profiler for Irregular Spreadsheets

An architecture note on recovering table structure from irregular workbooks, constraining model judgment to typed decisions, and giving deterministic validation the final write.

Spreadsheet intelligence · Typed operations · Model review · Deterministic validation · Lineage

Read architecture note
SYSTEM MAP: EVIDENCE_TO_VALIDATED_PROFILE
Four-phase spreadsheet profiler architecture: evidence, typed candidates, deterministic authority, and validated profiling with lineage.
Scroll the diagram or open the full-size searchable SVG.

Experience

[ TIMELINE ]
STATE RAIL
  1. 01
    –

    AI Engineer, LLM & Agentic Systems

    Saman Kavan Vidad

    Designed stateful agent workflows with structured contracts, validation gates, recovery strategies, human review, asynchronous execution, and artifact handling.

  2. 02
    –

    Data Analyst

    Atieh Dade Pardaz

    Built behavioral features and evaluated clustering strategies for stable, interpretable customer segments.

Capabilities

[ MATRIX ]
SKILLS INDEX
01

Agent systems

LangGraph · multi-agent orchestration · structured generation · human-in-the-loop workflows

02

Backend & data

FastAPI · Pydantic · JSON Schema · PostgreSQL · RabbitMQ · MinIO/S3

03

ML & analysis

Python · SQL · pandas · NumPy · PyTorch · scikit-learn · PyArrow

04

Engineering

Docker · asynchronous processing · workflow state · validation · event-driven architecture

Education

[ ACADEMIC ]
BACKGROUND
01

MSc, Computer Engineering · Computer Games

Iran University of Science and Technology

GPA
17.81/20
02

BSc, Computer Engineering · Software Engineering

Semnan University

GPA
17.45/20

Let’s talk about AI systems that need to hold up outside a demo.

I’m based in Tehran and reachable by email.

Email Mohadese GitHub
EMAIL mohadesehzamani1380@gmail.com