PROFILE

AI Engineer with 2+ years of experience, specializing in Computer Vision and actively transitioning to Generative AI. Proven ability to develop and deploy real-time video analytics (face recognition, crowd monitoring) and advanced GenAI solutions, including multi-agent LLM workflows (LangChain, LangGraph), RAG with ChromaDB, and knowledge graphs (Neo4j). Proficient in Python, SQL, and FastAPI, with a track record of building impactful prototypes such as intelligent chatbots, retrieval systems, and AI surveillance.

SKILLS
  • •Programming: Python, SQL
  • •LLM/Agents: LangChain, LangGraph, Agent routing, Tool/Function calling
  • •LLMs/Serving: Google Gemini 2.5 Flash, Qwen3-30B (Fireworks), LLaMA 3.2 (Ollama)
  • •RAG/Embeddings: ChromaDB, spaCy embeddings, mxbai-embed-large, retrieval pipelines
  • •Knowledge Graphs: Neo4j, Cypher, Entity extraction, Graph-RAG
  • •Apps/UI: Streamlit, Gradio
  • •Data/Backend: Pandas, SQLAlchemy, FastAPI, REST APIs
  • •Observability: LangSmith tracing and evaluation
  • •CV/ML: OpenCV, MTCNN, face_recognition, scikit-learn (DBSCAN/HDBSCAN/MeanShift/KMeans)
EXPERIENCE
AI Engineer, Xloop Digital Services
Nov 2022 – May 2025 | Karachi, PK
  • •Transitioned to GenAI: designed LLM-based chatbots/agents with retrieval over SQL and documents, vector search (ChromaDB), and tool-routed multi-agent workflows with LangGraph.
  • •Shipped prototypes with Streamlit/Gradio UIs and LangSmith tracing for observability and debugging.
  • •Built real-time Computer Vision systems: face recognition pipelines (MTCNN + face_recognition), crowd monitoring with clustering and severity thresholds, multi-stream threading, and SQL-backed event logging.
  • •Designed FastAPI-based RTSP streaming services; integrated JWT-secured REST, SQLAlchemy schemas, and robust logging/thresholding for production stability.
    • PROJECTS
    DataBot, Multi-Agent Chatbot with RAG and SQL Database | Xloop Digital Services
  • •Built a LangGraph-driven agentic system that routes user queries to SQL agents, RAG tools, or web search.
  • •Integrated conversation memory (daily CSV logs), monitoring (LangSmith), and a Streamlit UI.
  • •Tools: Primary agent with Fireworks Qwen3-30B; SQL agents with Gemini 2.5 Flash; ChromaDB + spaCy; Tavily search.
  • •Outcomes: robust multi-source Q&A covering structured/unstructured data; traceable executions and reduced failure modes via tool specialization.
    • VisionX – AI Surveillance System, Textile Sector | Xloop Digital Services
  • •Built crowd/incident detection using Ultralytics YOLO + clustering with severity state machine and debouncing.
  • •Designed event data model and migrations (SQLModel + Alembic); integrated auth (PyJWT), config (python-dotenv), and async services (FastAPI/Uvicorn).
  • •Delivered alerting for 3 monitored locations with 3 alert types.
  • •Implemented inference/runtime options with onnxruntime/torch and CV utilities (OpenCV, cvzone, filterpy).
  • •Annotations: 12k+ images labeled by me for this project.
    • EDUCATION
    B.E. Computer System Engineering, Dawood University of Engineering and Technology
    Nov 2018 – Nov 2022 | Karachi, PK
  • •CGPA: 3.71/4.00