Nazmi

Selected work

AI agents and models, shipped not slideware.

Representative work across client projects, products and systems, open source, and research.

Client projects

AI agent and model projects built with client teams.

Production AI systems built with client teams.

  1. Recommendations and an AI shopping assistant for an online fashion platform

    BuiltContributed to the recommendation systems and the conversational shopping assistant as part of the client team.

    ResultLive in production for customers.

  2. Production ML workflows for an enterprise telecom

    BuiltML workflows on Azure Machine Learning and Databricks.

    ResultStandardized and shipped into production.

  3. RAG customer-support assistant

    BuiltA retrieval-augmented (RAG) assistant, from prototype to production.

    ResultHandling first-line support, handed to the team to run.

  4. AIOps monitoring for a wireless network operator

    BuiltAnomaly detection on wireless and 5G data, LLM log processing, and root-cause retrieval.

    ResultShipped into network operations.

  5. Energy forecasting for a commercial-buildings firm

    BuiltConsumption forecasting models and streaming monitoring pipelines.

    ResultShipped forecasting and monitoring into production.

  6. Visual search and recommenders for a fashion marketplace

    BuiltImage similarity search over a million-scale catalog, with complementary-item recommendations.

    ResultShipped as a serving API behind the marketplace.

Products & systems

Earlier products and platforms.

Products and platforms Isaac helped build before founding Nazmi.

  1. Resoniks

    Industrial inspection: acoustic anomaly detection from architecture to production across hybrid cloud and on-prem, with sub-100ms inference.

    ResultShipped product, in production use.

  2. In Parallel

    Decision support: agentic systems for planning and monitoring, with event-driven multi-agent workflows and memory across sessions, graphs, and documents.

    ResultShipped agentic platform for planning and organizational learning.

  3. Silo AI (acquired by AMD)

    Autonomous driving: simulation and validation that cut test-iteration time from days to hours.

    ResultAdopted by customer engineering teams.

Open source

Code you can read.

A working artifact you can inspect and run.

  1. Lerim

    A context compiler for AI agents: turns finished agent traces into reusable, cited memory the next agent reads instead of starting cold.

    View on GitHub

Research & writing

Publications.

Research on machine learning, multi-agent systems, and autonomous driving.

  1. MACRPO: Multi-Agent Cooperative Recurrent Policy OptimizationFrontiers in Robotics and AI, 2024
  2. COSBO: Conservative Offline Simulation-Based Policy OptimizationarXiv, 2024
  3. KoMA-GIAM: A Graph-based Multi-agent Framework for Safety-Enhanced Autonomous Driving with LLMsUnder review, IEEE Transactions on Intelligent Vehicles
  4. Learning-Based High-Level Decision-Making for Abortable Overtaking in Autonomous VehiclesIEEE Transactions on Intelligent Transportation Systems, 2022
  5. Vision Transformer for Learning Driving Policies in Complex and Dynamic EnvironmentsIEEE Intelligent Vehicles Symposium, 2022
  6. Increasing the Efficiency of Policy Learning for Autonomous Vehicles by Multi-Task Representation LearningIEEE Transactions on Intelligent Vehicles, 2022
  7. Using an Encoder-Decoder Convolutional Neural Network to Predict the Solid Holdup Patterns in a Pseudo-2D Fluidized BedChemical Engineering Science, 2021

More writing on Medium →Full history on LinkedIn →

Start

Build your AI system with Nazmi.

Tell us what you are building, what exists today, and where your team needs help.

Start a conversation or book a 20-minute call →