I am an AI researcher and practitioner with over a decade of experience in artificial intelligence, specializing in large language models (LLMs), generative AI, agentic systems, and their transformative real-world applications. I have extensive experience architecting and deploying sophisticated AI solutions across various sectors, including Telecommunications, Retail, Finance, Healthcare, and Robotics.
Currently, I am the Founder of Nazmi, an independent AI consulting practice focused on helping enterprise clients design and build practical machine learning, LLM, and data-intensive AI systems. I also founded and maintain Lerim, an open-source context compiler for AI agent workflows that turns completed agent sessions into cited, reusable context records.
Previously, as Principal AI Scientist at In-Parallel, I led the design and implementation of advanced agentic AI systems utilizing graph-based Retrieval-Augmented Generation (Graph-RAG) and multi-agent systems to enhance strategic decision-making, planning, execution monitoring, and knowledge extraction capabilities. Before that, as Co-Founder and CIO/CAIO at Resoniks, I spearheaded the development of an AI-driven anomaly detection and quality control system utilizing acoustic data, and played a key role in securing EUR 2.65 million in seed funding.
My research has been published in top journals and conferences, including Frontiers in Robotics and AI (MACRPO), IEEE Transactions on Intelligent Vehicles (Multi-Task Representation Learning), and IEEE Intelligent Vehicles Symposium (Vision Transformers for Driving Policies). With a passion for innovation, I mentor aspiring AI professionals and have authored over 100 blog posts to make complex AI topics accessible.
Open-source context compiler for AI agent workflows. Turns completed agent sessions into cited, reusable context records.
AI-assisted monitoring workflows for wireless network operations: anomaly detection, log analysis, root cause analysis, and event correlation.
Enterprise customer support assistant using retrieval-augmented generation, semantic search, and production deployment workflows.
AI assistant for clinical documentation, medical search, and doctor-patient interaction using Neo4j knowledge graphs and RAG.
Machine learning components for energy forecasting and real-time monitoring in commercial buildings using Kafka and AWS.
Visual search and outfit recommendation for million-scale fashion images using PyTorch, FAISS vector search, and Flask APIs.
Sharing insights on AI, machine learning, and technology
Recently, I have started to play around and learn more about DSPy and decided to write several blog posts as I learn more. And as the medical domain is one of my interests, I thought we could do a simple project there...
Read more →As someone with a background in reinforcement learning (RL) and having witnessed its rising prominence in the large language model (LLM) domain, I have been thinking a lot over the last few weeks about how RL is used to refine LLM behavior. From fine-tuning policies ...
Read more →