Contact
Get in touch.
Whether you're prototyping a robot, integrating AI into your product, or need a professional website — get in touch. Available for consulting, contract engineering, web development, and research collaboration.
Robotics · AI · Web Applications
Robotics researcher, AI engineer, and full-stack web developer. I help startups, research labs, and companies build sophisticated technical products — robots, AI systems, and the web applications that power them.
What I Do
I bring the same rigorous engineering mindset to a robot that needs to navigate and adapt, an AI system integrated into your product, or a professional web presence that converts.
Reinforcement learning applied to multi-DOF arms, locomotion, and manipulation, built and validated in custom simulation environments.
End-to-end vision pipelines from camera hardware to detection algorithms, including field-deployed systems.
Hard real-time OS design and low-level firmware for robotic control hardware.
Applied ML engineering from research to production, built to integrate cleanly into a real product.
Custom physics simulation and data tooling for problems too large or too complex for off-the-shelf solutions.
Professional websites and AI-powered web applications, built clean and fast — no templates, no page builders.
About
Roman Aguilera is a robotics researcher and engineer whose work sits at the intersection of AI and physical systems — designing algorithms that give robots the ability to learn, adapt, and operate in complex, real-world environments.
His background spans the full stack: from embedded electronics and computer vision to reinforcement learning and real-time OS development. He's shipped research at academic and startup levels, and led AI engineering in a VC-backed product context.
If your team is building something that moves, perceives, or decides — Roman can help you do it rigorously and faster.
Experience
Projects
Research on applying reinforcement learning to high-DOF robotic control. Investigates whether RL develops true goal reasoning or trajectory heuristics. Achieved stable reaching policies up to 32 DOF using N-link manipulator simulators with PPO/TRPO/SAC algorithms.
View Blog Post →Addresses computational tractability for high-DOF robot controllers. Uses first-principles state-space bounding and Barycentric Interpolation to enable Value Iteration on dynamical systems without combinatorial blowup.
View on GitHub →End-to-end hard real-time control system for agile humanoids, integrating Xenomai & Ubuntu into a custom RTOS purpose-built for robotic applications.
View on GitHub →Tool to programmatically generate robot model XMLs for physics simulators including MuJoCo and PyBullet. Parameterized by link count, geometry, and joint config.
Tableau dashboards analyzing spatial advertising performance across CVR, CTR, CPC, CPM, and 5 other KPIs to identify high- and low-performing ad groups by geography.
View on Github →Data-driven recommendations for health violation prevention, identifying highest-risk establishments and advising targeted training interventions using NYC open data.
View on Github →Deep generative image model built from scratch in PyTorch, implementing recurrent pixel-by-pixel generation.
An overview on the theory, history, and applications of IBM's TrueNorth Neurosynaptic Procssor.
View on GitHub →A wireless health monitoring prototype to measure and track skin flexion information. Health information is sent to laptop and smartphone application. Used PSoC4 development board.
View on GitHub →Education
Contact
Whether you're prototyping a robot, integrating AI into your product, or need a professional website — get in touch. Available for consulting, contract engineering, web development, and research collaboration.