umer J.
Senior ML/AI Engineer
Open to offersNew to PlatformAbout
Umer J. is a highly skilled Senior ML/AI Engineer with over a decade of experience in deploying intelligent systems across varied sectors such as government, fintech, e-commerce, and healthcare. Operating primarily between Dubai and Islamabad, Umer specializes in LLM orchestration, Agentic AI architectures, and multi-agent workflows. With a talent for building open-source AI tools, particularly the Agentic graph RAG system, he has effectively managed operations at a national scale, notably supporting over 5 million users. Umer has spearheaded several AI deployments on platforms such as Azure AI Foundry, leveraging strong skills in PyTorch, TensorFlow, and frameworks like HuggingFace Transformers. His role at LogicEra involved architecting ML pipelines, improving RAG orchestration, and enhancing LLM fine-tuning methodologies. At Digital Dubai Authority, he directed AI projects like the Dubai AI Chatbot, augmenting infrastructure for digital identity platforms, and orchestrating sophisticated data storage solutions. Umer's commitment to technology extends to open-source contributions, specifically with high concurrency ML ingestion in Neo4j systems, showcasing his knack for engineering robust, innovative solutions.
Experience
Lead ML / AI Engineer
LogicEra – Microsoft Solutions Partner · 2024 — PresentArchitected and deployed production-grade ML and AI pipelines using Azure AI Foundry, ensuring optimal endpoint management across GPT-4o, Phi-3, and Mistral models. Designed Prompt Flow DAGs for enterprise RAG while assessing risks related to hallucinations and prompt injection attacks. Developed multi-agent network topologies via LangGraph orchestration, integrating LiteLLM dynamic model routing and Qdrant/Neo4j endpoints for structural knowledge retrieval. Fine-tuned LLMs on Arabic educational datasets to improve domain accuracy and implemented OCR sanitization for data ingestion. Created Edulytics, a solution that translates natural language queries into structured Cypher and SQL commands applicable to institutional databases.Lead AI Engineer & Senior Mobile Developer
Digital Dubai Authority · 2023 — 2024Oversaw architectural governance for the UAE Pass national digital identity platform, ensuring secure authentication for over 5 million users. Developed the conversational AI framework for the Dubai Government AI Assistant, incorporating a Model Context Protocol (MCP) to manage queries across various data sources, allowing for context-aware responses in both Arabic and English. Designed a multi-source RAG pipeline that aggregates government datasets and service catalogs, applying advanced retrieval methods to enhance reply precision for citizens. Architected a data storage system integrating vector databases, relational stores, and document repositories, creating ingestion pipelines with NLP/NER preprocessing to ensure data relevance. Maintained and published critical government mobile SDKs while implementing security standards for token encryption and runtime storage. Directed CI/CD quality assurance with tools like SonarQube and CodeQL to sustain high code coverage.Senior Systems & Application Engineer
Moove · 2022 — 2023Developed and enhanced micro-features for a transit platform designed to manage real-time telemetry from over 10,000 active drivers across various markets. Revamped the legacy CI/CD pipeline by integrating CircleCI, which significantly reduced manual handoffs and increased the efficiency of cross-team deployments.Core Creator & Maintainer
OPEN SOURCE: AGENTIC GRAPHRAG ENGINEEngineered a unified engine that combines structured transactional records and unstructured documents in a Neo4j database, breaking down traditional data silos. Developed an intelligent router for dynamic intent routing, classifying queries to either Cypher aggregation handlers or semantic graph searches. Executed high-concurrency data ingestion utilizing Redis queues, PyMuPDF for parsing, and multi-threaded NER processing, enabling bulk Neo4j writes at scale. Created a comprehensive Playwright regression suite ensuring precision across response validation scenarios.
Skills & Expertise
Education
- Bachelor of Science in Computer ScienceMuhammad Ali Jinnah University · 2009 — 2014