AI Engineer => from low-level neural hardware to high-level agent systems.
I build resource-efficient AI. Over years I've shipped automotive vision pipelines, optimized models for edge GPUs, and published research on spiking neural networks. Now I'm channeling that systems-level thinking into agentic AI and LLM tooling — designing frameworks where autonomous agents reason, plan, and act reliably.
🚧 New open-source project — coming soon.
An agentic framework and set of libraries for building with large language models. Details, architecture docs, and initial code drops are on the way. Star or watch this repo to follow along.
My path through AI has been bottom-up: starting at the neuron level and working toward full autonomous systems.
Neuromorphic Computing — Designed energy-efficient spiking neural networks for classification and robotic control using surrogate gradient methods and SNNTorch. Published research on SNN-based policy learning for manipulation tasks.
Automotive Vision — Built end-to-end detection pipelines with YOLO-World and Swin-Transformer, deployed on NVIDIA Jetson edge hardware via Kafka-based streaming architectures.
Multi-Agent Reinforcement Learning — Developed transformer-based multi-agent RL systems at TRUMPF for industrial automation, achieving measurable gains in task completion speed.
Signal Processing — Placed in two IEEE hackathons building EEG-driven systems: emotion decoding with generative models and a brain-controlled driving interface.
-
Parametric Study for Lightweight Monocular Depth Estimation Deep Neural Network Comparative analysis of lightweight depth estimation on resource-constrained devices. Google Scholar
-
Policy Learning with Spiking Neural Networks for a Robot Manipulation Task Energy-efficient SNNs for robotic manipulation in neuromorphic environments. Google Scholar
Languages <=> Python, C++, CUDA
ML & DL <=> PyTorch, TensorFlow, PyTorch Lightning, Hugging Face Transformers, SNNTorch, OpenCV
Infrastructure <=> Docker, CUDA, ROS, Apache Kafka, MLflow, NVIDIA Jetson
Domains <=> Deep Learning, Computer Vision, NLP, Neuromorphic Computing, Reinforcement Learning, Multi-Agent Systems, Generative AI
| Project | What it does | Stack |
|---|---|---|
| Smart Vision Systems | Real-time automotive detection on edge GPUs | PyTorch, Kafka, Docker |
| Accent Analyzer | English accent detection from video | Streamlit, Azure Speech, FFmpeg |
Open-source contributions: Hugging Face Transformers (3 PRs merged), OpenCV-Python tutorials.