Adeeb Waizbuilds full-stack productswith AI at the core,from interface to infrastructure.

Full-stack engineer in Dubai. Three years of .NET and MERN, now RAG pipelines, agent orchestration and LLM guardrails.

Selected work

NL2SQL Engine (opens in a new tab)

2024

“Which regions missed target last quarter?”

Business users ask the data warehouse in plain English. The model writes SQL, and guardrails validate it against the schema before anything runs.

Built with .NET, Azure OpenAI and Angular.

Open the live app

SELECT region,
       SUM(revenue) - SUM(target) AS gap
FROM   quarterly_revenue
WHERE  quarter = '2024-Q3'
GROUP  BY region
HAVING SUM(revenue) < SUM(target);

✓ schema validated   ✓ read-only role

SELECT region,
       SUM(revenue) - SUM(target) AS gap
FROM   quarterly_revenue
WHERE  quarter = '2024-Q3'
GROUP  BY region
HAVING SUM(revenue) < SUM(target);

✓ schema validated   ✓ read-only role

60% fewer ad-hoc data tickets

The NL2SQL interface: a plain-English query, the generated SQL, and a results table

RAG Document Agent

2024

“What does the contract say about early termination?”

On-premise search over internal documents, answered by local LLMs with citations back to the source clause.

Built with Python, LangChain and pgvector.

embed(query, all-MiniLM-L6-v2)   → 384-dim vector
search(pgvector, k=8)            → 8 chunks
answer(local LLM, cite=true)     → §14.2, §14.3

✓ no data left the network

embed(query, all-MiniLM-L6-v2)   → 384-dim vector
search(pgvector, k=8)            → 8 chunks
answer(local LLM, cite=true)     → §14.2, §14.3

✓ no data left the network

500+ documents, zero external data leakage

Auto Data Agent

2023

“Collect this week's listings and clean them up.”

A collection pipeline that scrapes, normalizes and deduplicates records, then hands them to reporting workflows.

Built with Python, Playwright and n8n.

crawl(sources, workers=8)        → 10,247 records
normalize(schema="listing")      → typed rows
dedupe(key="source_id")          → 9,891 unique
trigger(n8n, "weekly_report_pipeline")

crawl(sources, workers=8)        → 10,247 records
normalize(schema="listing")      → typed rows
dedupe(key="source_id")          → 9,891 unique
trigger(n8n, "weekly_report_pipeline")

10,000+ records a week, no manual scraping

About

Full-stack engineer turned AI systems builder. Three years shipping .NET and MERN applications, now focused on production AI: RAG pipelines, agent orchestration, and LLM guardrails that hold up under enterprise scrutiny.

Recently I eliminated 20+ hours of manual reporting per cycle with an event-driven document pipeline, cut Azure OpenAI costs by 30%, and enforced data governance upstream.

Experience

2025 – nowFull Stack & Applied AI EngineerMarket-i
2023 – 2025Full Stack DeveloperINTELPEEK
2022 – 2023Web Developer InternVAI Marketing