About me

I am a Research Scientist at DeepMind, working on Function calling with LLMs, Exploration in Reinforcement Learning and Agent-Task co-evolution approaches.

Research Interests

Intersection of artificial intelligence and robotics.
- Representation learning
- Deep reinforcement learning
- Neuroevolution
- Quality-Diversity
- Robotics

Bio



I obtained my BSc degree in Mechatronics Robotics and Automation, at the Faculty of Technical Sciences, University of Novi Sad in 2011. Afterwards, I completed the double-degree EMARO (European Masters on Advanced Robotics) program and was awarded the MSc degree in 2013.

During 2013-2014, I worked as a research engineer in RIS group, LAAS-CNRS Toulouse on rover locomotion diagnostics using sequential machine learning models.

In 2015-2016, I was as a research assistant in the iBug group, Dept. of Computing, Imperial College London. There I worked on applying deep learning methods for human emotion recognition based on multimodal data, i.e. facial expressions and speech.

I started my PhD studies in 2016 in Robot Intelligence Lab, Imperial College London, under the supervision of Dr Petar Kormushev. During my PhD I interned at DeepMind at the Robotics Lab, hosted by Francesco Nori. I completed my PhD in 2021, with the thesis titled "Parameter space abstractions for diversity-based policy search".

Publications



2023


  • Gemini: A family of highly capable multimodal models

    Gemini Team
    ArXiV

    [pdf] [citation] [blog]


  • Human-timescale adaptation in an open-ended task space

    AdA team: Jakob Bauer, Kate Baumli, Feryal Behbahani, Avishkar Bhoopchand, Nathalie Bradley-Schmieg, Michael Chang, Natalie Clay, Adrian Collister, Vibhavari Dasagi, Lucy Gonzalez, Karol Gregor, Edward Hughes, Sheleem Kashem, Maria Loks-Thompson, Hannah Openshaw, Jack Parker-Holder, Shreya Pathak, Nicolas Perez-Nieves, Nemanja Rakicevic, Tim Rocktäschel, Yannick Schroecker, Satinder Singh, Jakub Sygnowski, Karl Tuyls, Sarah York, Alexander Zacherl, Lei M Zhang
    International Conference on Machine Learning
    (Oral presentation)

    [pdf] [citation] [website] [New Scientist]




2021


  • Policy Manifold Search: Exploring the Manifold Hypothesis for Diversity-based Neuroevolution

    N. Rakicevic, A. Cully and P. Kormushev.
    Genetic and Evolutionary Computation Conference, 2012

    [pdf] [citation] [website] [video] [code]


  • ResQbot 2.0: An Improved Design of a Mobile Rescue Robot with an Inflatable Neck Securing Device for Safe Casualty Extraction

    R.P. Saputra, N. Rakicevic, I. Kuder, J. Bilsdorfer, A. Gough, A. Dakin, E. de Cocker, S. Rock, R. Harpin and P. Kormushev.
    Multidisciplinary Digital Publishing Institute (MDPI), 2021

    [pdf] [citation] [website]


  • Hierarchical Decomposed-Objective Model Predictive Control for Autonomous Casualty Extraction

    R.P. Saputra, N. Rakicevic, D Chappell, K Wang and P. Kormushev.
    IEEE Access, 2021

    [pdf] [citation] [website]




2020


  • Policy Manifold Search for Improving Diversity-based Neuroevolution

    N. Rakicevic, A. Cully and P. Kormushev.
    Beyond Backpropagation: Novel Ideas for Training Neural Architectures Workshop, (NeurIPS'20)
    [oral 8% acceptance rate]

    [pdf] [citation] [video] [poster]




2019


  • Sim-to-Real Learning for Casualty Detection from Ground Projected Point Cloud Data

    R.P. Saputra, N. Rakicevic and P. Kormushev.
    International Conference on Intelligent Robots and Systems (IROS), 2019

    [pdf] [citation] [website]


  • Active Learning via Informed Search in Movement Parameter Space for Efficient Robot Task Learning and Transfer

    N. Rakicevic and P. Kormushev.
    Autonomous Robots (AURO), 2019

    [pdf] [citation] [website] [code]




2017


  • Efficient Robot Task Learning and Transfer via Informed Search in Movement Parameter Space

    N. Rakicevic and P. Kormushev.
    Workshop on Acting and Interacting in the Real World: Challenges in Robot Learning (AIRW), (NIPS'17)

    [pdf] [citation] [poster]




2016


  • Multi-modal Neural Conditional Ordinal Random Fields for Agreement Level Estimation

    N. Rakicevic, O. Rudovic, S. Petridis and M. Pantic.
    International Conference on Pattern Recognition (ICPR'16)

    [pdf] [citation]




2015


  • Neural Conditional Ordinal Random Fields for Agreement Level Estimation

    N. Rakicevic, O. Rudovic, S. Petridis and M. Pantic.
    1st International Workshop on Automatic Sentiment Analysis in the Wild (WASA'15), (ACII'15)

    [pdf] [citation]