Available for AI engineering roles · Dubai, UAE

I build AI that ships:
LLM apps, agents and RAG
running in production.

I'm Mohammad Hayat, an AI/ML Engineer and Generative AI Solutions Architect. For 3 years I've designed and deployed LLM platforms on Microsoft Azure and Google Cloud, from the architecture through evals, guardrails and monitoring.

See my work Download CV Get in touch
  • 3 yrsproduction GenAI
  • 10+AI projects delivered
  • 14M+ad impressions on markXpand
  • ~60%less manual ESG tagging

01 / Selected work

Live products I've built and engineered

Each one is running in production. Click through to see it.

PersonaXpand homepage: post to 8 social networks from one place

PersonaXpand

Web · iOS · Android

AI social media manager · built end to end

  • Write once and publish to 8 networks, with each version adapted to the platform's limits, formats and hashtags.
  • An LLM Digital Twin learns the user's tone and topics, then writes their captions, hashtags, images and replies.
  • AI auto-reply to comments, DMs and mentions, with human-in-the-loop approval and loop guardrails.
  • OAuth with AES-256 encrypted tokens, quota-aware publishing, circuit breakers and GDPR/CCPA data controls.
personaxpand.com ↗
markXpand homepage: AI-powered marketing platform

markXpand

Live

AI marketing platform

  • Turns products and goals into live campaigns on Google, Meta, YouTube, TikTok, LinkedIn, Snapchat, email and WhatsApp.
  • Competitor research from the live web, budget allocation, and Brand Kit extraction that feeds on-brand copy, images and video.
  • 14M+ impressions and 3.7M+ people reached for 15+ brands, per publicly reported figures.
markxpand.com ↗
PurchoSmart brand pattern

PurchoSmart

Live

AI procurement voice agent · developer

  • An AI voice agent calls up to 20 UAE suppliers at once and collects live quotes.
  • Speaks 5 languages (English, Arabic, Hindi, Tamil, Spanish) and follows up on WhatsApp and email.
  • Returns a ranked report on price, delivery and availability, with recordings and transcripts, in about 12 minutes.
purchosmart.com ↗
ReachXpand homepage: your business, growing on autopilot

ReachXpand

Live

AI social automation for businesses

  • Multi-tenant: each brand has its own workspace, Brand Kit and AI voice across 8 networks.
  • A grounded auto-reply answers only from approved brand facts, never invents answers, and hands unknown questions to a human.
  • Brief → strategy → storyboard → rendered video for Reels, TikTok and Shorts. The inbox also separates leads from complaints.
reachxpand.com ↗
bodymaths homepage: a free fitness coach

bodymaths

Live

Free fitness and nutrition planner · personal project

  • An on-device camera rep counter: video never leaves the phone.
  • Meal logging from a photo or a short description, 3D exercise demos, and a goal forecast date.
  • No signup and no ads, plus 38 free calculators. Built to cost nothing to run.
bodymaths.com ↗

Brisky & ESG XBRL engine

BriskFlow

RAG for sustainability reporting

  • A RAG engine grounded in IFRS, ESRS and SASB taxonomies that predicts and validates XBRL tags.
  • Brisky is an ESG chatbot built with a LangChain agent and Gemini on Vertex AI, and it explains its reasoning.
  • Tuned chunking, embeddings and re-ranking over FAISS and ChromaDB cut manual tagging by about 60%.

Python · LangChain · Vertex AI · FastAPI · FAISS · ChromaDB · Docker

More projects

AI & machine learning

Agentic Chatbot

A multi-agent chatbot built with CrewAI and DeepSeek. It keeps conversation memory for follow-ups and runs on a FastAPI backend.

CrewAI · DeepSeek · FastAPI

Private repo (BriskFlow)

Financial Analyst with DeepSeek

CrewAI agents read a stock query, run DeepSeek on SambaNova and write an analysis summary. The interface is a Streamlit app.

CrewAI · DeepSeek · SambaNova · Streamlit

GitHub ↗

Plant Disease Prediction

A CNN image classifier trained on the PlantVillage dataset. It's packaged as a Dockerized web app, with a GitHub Actions image build.

TensorFlow · CNN · Docker · GitHub Actions

GitHub ↗

Netflix Recommendation System

A content-based recommender that matches titles by TF-IDF text vectors and cosine similarity.

Python · scikit-learn · TF-IDF

GitHub ↗

Data analytics

Customer Acquisition Cost Analysis

Compares acquisition cost and return across marketing channels to show where the budget works best.

Python · Pandas · Plotly

GitHub ↗

Retail Sales Analysis

SQL for cleaning and exploring retail sales data, covering sales trends and customer behaviour.

SQL · EDA

GitHub ↗

Data Professional Survey Dashboard

A Power BI dashboard of a global survey of data professionals, covering salaries, job satisfaction and roles.

Power BI · Data viz

GitHub ↗

Food Delivery Cost & Profitability

Breaks down delivery fees, discounts and commissions per order to find where costs can drop and margins grow.

Python · Pandas

GitHub ↗

Forbes Billionaires Analysis

Explores Forbes billionaire data by industry, country and demographics.

Python · Data analysis

GitHub ↗

02 / Experience

Where I've worked

  1. Jan 2026 to presentDubai, UAE

    AI / ML Engineer · Nextoria Info Tech

    • Built PersonaXpand end to end, and engineered AI across markXpand, ReachXpand and PurchoSmart.
    • Set the AI architecture for enterprise GenAI on Azure AI Foundry and Azure OpenAI (GPT-4o, GPT-4.1).
    • Built agentic RAG with Azure AI Search (hybrid vector and keyword search with semantic ranking) and Document Intelligence ingestion.
    • Built multi-agent workflows with Semantic Kernel and Foundry Agent Service, using tool calling and structured outputs.
    • Ran LLMOps: prompt versioning, groundedness evals, Content Safety guardrails and Application Insights monitoring.
    • Led solution design sessions with stakeholders and compared models and vendors on cost, latency and accuracy.

    Python · C# · .NET · Azure AI Foundry · Azure OpenAI · AI Search · Semantic Kernel · Container Apps · Functions · Azure DevOps

  2. Jan 2024 to Jan 2026BriskFlow

    AI Developer · BriskFlow

    • Built an AI ESG XBRL tagging engine using NLP and RAG across the IFRS, ESRS and SASB standards.
    • Built Brisky, an LLM ESG chatbot that uses LangChain agents and Gemini on Vertex AI.
    • Built document parsing and auto-recommendation that cut manual annotation time by about 60%.
    • Shipped a Dockerized FastAPI backend with structured outputs, so tagging decisions can be audited.

    Python · LangChain · Vertex AI (Gemini) · FastAPI · Pydantic · FAISS · ChromaDB · Docker

03 / Skills

What I work with

Generative & agentic AI

LLMs · RAG · Agentic RAG · Hybrid search · Re-ranking · Multi-agent systems · MCP · Tool calling · Structured outputs · Fine-tuning (LoRA / PEFT)

Microsoft Azure AI

AI Foundry · Azure OpenAI · Foundry Agent Service · AI Search · Document Intelligence · Content Safety · Prompt Flow · Azure ML

Google Cloud AI

Vertex AI · Gemini 2.5 Pro / Flash · Vertex AI Agent Builder · Cloud Run · Cloud Storage

Frameworks

LangChain · LangGraph · LlamaIndex · Semantic Kernel · CrewAI · AutoGen · Hugging Face · FastAPI · PyTorch · TensorFlow · scikit-learn

LLMOps & evaluation

RAGAS · Azure AI Evaluation · LangSmith · Prompt versioning · Guardrails · Tracing · Cost & latency tuning · Responsible AI

Data & vectors

Azure AI Search · FAISS · ChromaDB · pgvector · Embeddings · Chunking · OCR · NLP · SQL Server · PostgreSQL

Cloud & DevOps

Functions · App Service · Container Apps · AKS · Cosmos DB · Key Vault · API Management · Docker · Kubernetes · GitHub Actions · Azure DevOps

Languages

Python · C# · .NET · SQL

Education

Master of Computer Applications (MCA)Islamic University of Science & Technology, Kashmir · 2021 to 2023
Bachelor of Computer Applications (BCA)Amar Singh College, Srinagar · 2017 to 2020

04 / Contact

Building something with LLMs?
Let's talk.

I'm open to AI/ML engineering and GenAI architecture roles, plus consulting.

Emailhayat28mir@gmail.com Phone · WhatsApp+971 50 508 3587 LinkedInin/hayatm GitHubhayat-personal