Skip to content

Experience

Four years of production AI.

From ML pipelines and fine-tuning to multi-agent architecture and banking cloud platforms for UAE and GCC clients.

  1. Nov 2024 - Present

    Lahore, Pakistan

    AI / ML Engineer & AI Solution Architect

    DigiFloat

    Architect and ship enterprise Generative AI for UAE energy, banking and government clients. Built 25+ agentic AI systems on LangChain, LangGraph, Azure AI Services and Copilot Studio, and architect platforms end to end across event-driven, microservices and modern-monolith designs deployed on AKS with Helm charts.

    • Built 25+ agentic AI systems for enterprise Law, Data and UAE/GCC clients using LangChain, LangGraph, Azure AI Services and Copilot Studio
    • Pioneered production use of MCP (Model Context Protocol) and A2A (Agent-to-Agent) for secure, interoperable multi-agent systems
    • Set up MLOps on Azure DevOps with CI/CD and model versioning; deployed across Container Apps, Functions and AI Foundry, with Langfuse for observability
    • Delivered USD 50M+ revenue impact and 1000+ engineering hours saved annually via the Data2Decisions multi-agent platform
    • Cut multilingual RAG latency 73% (15s to 4s) for 1000+ government users on RTA Mahboob, with text-to-SQL over 500+ tables
    • Reached 95% extraction and 98.75% evaluation accuracy on the Fertiglobe Microsoft Fabric policy extraction pipeline
    • Python
    • FastAPI
    • LangChain
    • LangGraph
    • CrewAI
    • Azure AI Foundry
    • Microsoft Copilot Studio
    • Azure OpenAI
    • MCP
    • A2A
    • Microsoft Fabric
    • Apache Kafka
    • Azure Event Hubs
    • AKS
    • Helm
    • Azure Container Apps
    • Azure Functions
    • Azure AI Search
    • PostgreSQL
    • Redis
    • Azure DevOps
    • Langfuse
    • Microsoft Entra ID

  2. Jun 2022 - Present

    Remote

    AI Solution Architect, AI Consultant & ML Engineer

    Vector AI Lab

    AI solution architect and consultant for enterprise clients: advising on AI strategy, designing target architectures as modern (modular) monoliths or microservices, and delivering AI/ML, agentic AI and data science solutions across NLP, computer vision and predictive modelling. Also building the lab's own products, Vector-Hub and VectorLumira.

    • Led product design and solution architecture for Vector-Hub (vector-hub.org), a connector hub that plugs AI agents into enterprise tools and data sources
    • Built VectorLumira (vectorlumira.com), a platform for building enterprise AI agents and workflows
    • Consulted with clients from discovery to delivery: requirements workshops, AI roadmaps and solution designs
    • Architected client platforms as modern (modular) monoliths or microservices, choosing per workload on scale, cost and team size
    • Delivered 30+ AI/ML projects across NLP, computer vision and predictive modelling
    • Built and deployed 10+ agentic AI, RAG and fine-tuning pipelines for diverse client use cases
    • Engineered NLP solutions including text classification, sentiment analysis and conversational agents
    • Developed computer vision systems for image recognition and real-time object detection
    • Designed automated ETL pipelines, improving client operational efficiency by 60%

    Products

    • Solution Architecture
    • Microservices
    • Modular Monolith
    • AI Consulting
    • Python
    • LangChain
    • LangGraph
    • CrewAI
    • Agentic AI
    • PyTorch
    • TensorFlow
    • Hugging Face
    • OpenAI
    • FastAPI
    • MongoDB
    • Docker
    • CI/CD Pipelines
    • Azure
    • AWS
    • GCP

  3. Jan 2024 - Jul 2024

    Remote

    Data Scientist / ML Engineer

    Swift Solver

    Built end-to-end ML pipelines across 50K+ records and fine-tuned small language models with LoRA/QLoRA, alongside NLP classification and sentiment pipelines with automated retraining.

    • Built end-to-end ML pipelines across 50K+ records
    • Fine-tuned SLMs with LoRA/QLoRA for a 25% accuracy gain
    • Built NLP classification and sentiment pipelines using Transformers and BERT
    • Set up model versioning, experiment tracking and automated retraining
    • Python
    • Transformers
    • BERT
    • Hugging Face
    • PyTorch
    • LoRA
    • QLoRA
    • PEFT
    • PySpark
    • ETL Pipelines
    • MLflow

  4. Jun 2023 - Dec 2023

    Lahore, Pakistan

    Python Developer / AI Engineer

    Pure Logics

    Fine-tuned GPT-3.5-turbo via the OpenAI API for a 40% accuracy gain and built preprocessing pipelines across 100K+ documents, shipping FastAPI and Streamlit apps to production.

    • Fine-tuned GPT-3.5-turbo via OpenAI API with a 40% accuracy gain
    • Built preprocessing pipelines across 100K+ documents
    • Shipped FastAPI and Streamlit apps in production with auth, rate limits and logging
    • Applied few-shot and chain-of-thought prompting to lift task accuracy
    • Python
    • OpenAI GPT-3.5-turbo
    • FastAPI
    • Streamlit
    • Hugging Face
    • NLTK
    • Prompt Engineering