UsamaShakeel

AI engineer

Vertekx • Temps plein

May 2025 - Present • 1 yr 5 mos

Building AI systems that drive business value and innovation Designed and implemented end-to-end AI solutions covering LLM integration, AI automation, evaluation, and production deployment to solve real business problems. Developed AI-as-a-Service (AIaaS) workflows and scalable intelligent systems that convert complex requirements into reliable,production-ready AI applications. Integrated and fine-tuned large language models with prompt engineering and domain adaptation to improve task-specific accuracy, reliability, and response quality. Built agent-based AI workflows for tool use, API orchestration, and multi-step task automation across business processes. Built secure document intelligence and OCR-driven processing workflows to extract, structure, and validate information from unstructured business documents. Evaluated LLM, NLP, and automation systems using task-appropriate metrics, error analysis, and iterative optimization to improve production performance. Integrated AI models into production systems through APIs, ensuring reliable interaction between models and application components.

AI Developer

EnDevSols LLC • Temps plein

May 2023 - Jul 2024 • 1 yr 2 mos

Machine Learning | Deep Learning | LLMs | RAG | Computer Vision | NLP Designed and implemented end-to-end AI solutions covering data pipelines, model training, and deployment workflows to solve real business problems. Collected, cleaned, and preprocessed structured and unstructured datasets, including text, images, and documents, to ensure high-quality model input. Implemented and experimented with machine learning and deep learning algorithms across classical ML, neural networks,transformers, and vision-language models, including predictive analytics for forecasting, anomaly detection, classification, and recommendation systems trained on historical business data. Integrated and performed custom fine-tuning (including LoRA/PEFT) on foundation and domain models — OpenAI GPT-4, LLama,Google Gemini, and open-source LLMs — to improve task-specific accuracy, precision, reliability, and response quality. Designed and deployed Retrieval-Augmented Generation (RAG) pipelines with LangChain and vector databases (Qdrant, Pinecone etc) to ground AI outputs in verified business knowledge bases. Evaluated machine learning, deep learning, LLM, RAG, computer vision, and NLP models using task-appropriate metrics, error analysis, confidence thresholds, and iterative hyperparameter tuning to optimize production performance. Integrated AI models into production systems through APIs and microservices, ensuring reliable interaction between models and application components.