I'm Jalalledin "Moji" Taavoni — freelance AI Integration Architect and Data Engineer based in Milano 🇮🇹.
I take AI from notebook demos to systems that run: reliably, observably, at the right cost, sometimes on the edge. The hard part isn't the model — it's the integration: stable orchestration, eval gates that fire before users notice, guardrails that hold, and inference that can run on a laptop when the cloud isn't an option.
const moji = {
role: ["AI Integration Architect", "Data Engineer", "DataOps Consultant"],
stack: ["Azure", "Databricks", "LangChain", "PromptFlow", "Foundry Local",
"SQL Server", "Synapse", "Fabric", "dbt", "Neo4j"],
philosophy: "Thoughtful before fancy.",
education: "Computer Science + Digital Humanities · Università di Pisa",
currently: "Building production LLM systems & agents on Azure",
open_to: "Freelance projects · IT and Remote EU",
reach: ["mojitmj.github.io", "linkedin.com/in/mojitmj", "t.me/mojitmj"],
};☁️ Cloud · Platform · DevOps
🧠 AI · ML · LLMs
💾 Data Engineering · DataOps
💻 Languages · Tools
|
PowerShell tool that x-rays a SQL Server / Azure SQL instance in one command — full DDL, DMVs, backup history, security audit, design-quality checks, per-table data samples. Cross-platform schedulers (Task Scheduler · SQL Agent · SSIS · cron · systemd).
|
GitHub Actions workflow for Azure Data Factory — JSON schema validation, hardcoded-secret scanning, Key Vault enforcement, trigger-state gates on every PR. Drop-in for any ADF estate.
|
|
Digital-humanities side project: 175 years of Italian academies as a property graph in Neo4j, visualized in the browser with popoto.js. Where data engineering meets the archive.
|
Live portfolio: dual-positioning landing page (AI / DataOps / DE / BI / DA), animated streaming-source boot, EN/IT toggle with Italian-flag theme, live chat overlay, full visitor metadata pipeline.
|
From: 12 May 2026 - To: 19 May 2026
Total Time: 42 hrs 25 mins
JSON 14 hrs 3 mins ███████▓░░░░░░░░░░░░░░░░░ 30.40 %
Markdown 7 hrs 36 mins ████░░░░░░░░░░░░░░░░░░░░░ 16.45 %
PowerShell 6 hrs 43 mins ███▓░░░░░░░░░░░░░░░░░░░░░ 14.55 %
Text 6 hrs 33 mins ███▓░░░░░░░░░░░░░░░░░░░░░ 14.18 %
SQL 2 hrs 8 mins █░░░░░░░░░░░░░░░░░░░░░░░░ 04.63 %
YAML 1 hr 9 mins ▓░░░░░░░░░░░░░░░░░░░░░░░░ 02.52 %
XML 1 hr 6 mins ▓░░░░░░░░░░░░░░░░░░░░░░░░ 02.40 %
- 🔒 Closed issue #1 in mojiTMJ/mojiTMJ
- [Day 7 - Dense Embedding - RAG](https://dev.to/indumathi__r/day-7-dense-embedding-rag-ojh) Thu May 21 2026 3:52 AM- [Why teachers need explainable AI, not just accurate AI — building the KC dashboard](https://dev.to/orieken/why-teachers-need-explainable-ai-not-just-accurate-ai-building-the-kc-dashboard-4iga) Thu May 21 2026 3:51 AM- [Closing the feedback loop: how mistake classification drives adaptive problem selection in NumPath](https://dev.to/orieken/closing-the-feedback-loop-how-mistake-classification-drives-adaptive-problem-selection-in-numpath-5ce9) Thu May 21 2026 3:50 AM- [Amazon Quick: AWS's Agentic Workspace, Explained for Engineers](https://dev.to/jubinsoni/amazon-quick-awss-agentic-workspace-explained-for-engineers-3lc2) Thu May 21 2026 3:47 AM- ["My Coding Agent Remembered Sessions, Not Work. That Was the Bug"](https://dev.to/codekingai/my-coding-agent-remembered-sessions-not-work-that-was-the-bug-1n3e) Thu May 21 2026 3:46 AM
- 🤖 Production AI — taking LLM demos / RAG / agent prototypes to systems that survive Tuesday morning
- 🛡️ AI evaluation & guardrails — golden sets, drift detection, regression gates, jailbreak hardening
- ⚡ Edge AI — Azure AI Foundry Local · ONNX · on-device LLMs for latency- or privacy-bound workloads
- 🏗️ DataOps / Data platform — lakehouse design on ADF + Databricks, CI/CD, governance, FinOps
- 🔧 SQL Server modernization — legacy → Azure SQL / MI / Fabric with replayable migrations
- 📊 BI / Power BI rescues — slow reports, wrong numbers, ungoverned sprawl
shipping: production AI integrations on Azure for SMB clients in IT/EU
building: sqlsnapshot v2 with Azure SQL DB + Fabric warehouse coverage
exploring: on-device LLMs (Phi-3, Llama-3) via Foundry Local + ONNX
reading: "Designing Data-Intensive Applications" (annual re-read)
learning: proper formal verification for AI agents
sipping: a long espresso ☕
"Pipelines should be like good plumbing — invisible, reliable, and someone else's problem when they break. I'm the someone else."
🟢🤍🔴 Fatto con espresso ☕ & 🍕 a Milano · Ciao mondo!