Resume

CV

Timothy Kevin Johnsen

CONTACT

https://timkjohnsen.com | tjohnsen@uci.edu | https://linkedin.com/in/tim-johnsen

https://scholar.google.com/citations?user=nvidNbAAAAAJ&hl=en | https://github.com/wreckittim

SUMMARY

PhD Candidate (June 2026) in Computational Science specializing in Machine Learning (ML), Reinforcement Learning (RL), Agentic AI, Autonomous Vehicles, Autonomous Science, and Adaptive Neural Network Frameworks. Proven track record in developing full autonomy stacks, closed loop systems, self-selecting policies, long-horizon planning, action generation, reasoning frameworks, and predictive models as supported by eight peer-reviewed publications across UCI, NASA, LBNL, and industry. Expert in optimizing compute-heavy pipelines for resource constrained devices via knowledge distillation, scalable computing, dynamic sensing, and edge computing. Experienced in high performance computing (HPC) and training large-scale multimodal models. Foundational background in STEM and the physical sciences, emphasizing rigorous experimental design for iterative model improvement.

SKILLS

Deep RL (TD3, DQN), Deep ML, Knowledge Distillation, Multimodal Feature Extraction, Large Language Models (LLM) (Gemma3), Parallel Computing, Diffusion Policies.
In Situ Autonomous Science, Sensor Fusion, Uncertainty Analysis, Monte Carlo Sampling, Denoising and Gap-Filling, Data Processing, Spectral Analysis.
Autonomous Vehicles and Systems, Navigation, Path Planning, Collision Avoidance, Imitation Learning, Hierarchical Learning, Control, Visual-Language-Action (VLA).
Hardware-in-the-Loop, Sim-to-Real, Pre-Training, Post-Traning , Human-in-the-Loop
Transformers, CNNs, MLPs, GCNs, U-Nets, RNNs, Slimmable/Gated Networks, Split Computing, Spatio-Temporal, ARIMA, Finite Differencing Schemes, AutoEncoders
Python, C++, Java, PyTorch, StableBaselines3, Linux/Bash, SQL, Tensorflow.
HPC, SLURM, NVIDIA Jetson, AWS, Databricks, Microsoft AirSim, ROS 2, DJI API.

RESEARCH & PROFESSIONAL EXPERIENCE

PhD Researcher | Intelligent and Autonomous Systems Lab, UC Irvine | 2021 – Present

Adaptive Models: Experiment with RL-based policies (TD3, DQN) that autonomously scale model complexity and input modalities in response to context, at inference time, by optimizing the trade-offs between test-time compute, power needs, and task accuracy.
Modular Pipelines: Iteratively train embedded slimmable and multi-branch split gated networks via RL and knowledge distillation to implement computationally complex models on real-world, resource-constrained hardware (NVIDIA Jetson).
Agentic AI: Design hierarchical VLA frameworks leveraging Diffusion Policies for high-level waypoint generation and planning, and RL for low-level robot embodiment actions, in dynamic and unknown environments.
Dynamic Flow of Information: Pre-training from scratch and post-training with transfer learning of RGB-D encoders for downstream tasks, using CNNs, MLPs, transformers.
Testbed Environments: Train and evaluate in high-fidelity simulators (AirSim, CARLA, Unreal Engine) and deploy to real-world hardware (NVIDIA Jetson, ROS 2, DJI Tello)
Data Collection: generate real-world and simulated datasets connecting multimodal sensor observations, environments, optimal actions, and shortest paths (A*).
Map Building: process spatio-temporal sensors to refine 2D and 3D information.
Scaling Experiments: Build and routinely maintain AI-accelerated servers for distributed deep learning; streamlined data processing via Python/Linux automation.

Research Scientist | Lunar and Planetary Lab, University of Arizona (UoA) | Summer 2025

Multimodal Sensor Fusion: Rigorously architected an end-to-end pipeline in PyTorch integrating Raman spectra and images via data processing and CNN/MLP feature extractors for autonomous Martian and Lunar mineral classification.
Distributed Training: Effectively leveraged HPC and SLURM to manage large-scale planetary datasets (20+ years of heterogeneous data), optimizing downstream tasks through frozen-stem transfer learning.

Student Research Scientist | Scientific Data Division, LBNL | 2023 – 2024

Probabilistic Modeling: Implemented Monte Carlo Dropout and sensitivity analysis for uncertainty quantification, for interpretability in predictions and sensor placement.
Denoising Methods: Gap-filled and reduced inherent noise in real-world data by utilizing dropout and implicit/explicit Denoising AutoEncoder (DAE) techniques.
Spatio-Temporal Modeling: Developed a U-Net architecture to fuse satellite images, weather data, and soil sensors for large-scale watershed modeling.
Distributed Training: Executed large-scale parallel processing on NERSC HPC with SLURM and Python to accelerate big data processing and deep learning.

Intern Data Scientist | eHealth, Inc. | Summers 2020 – 2021

Human-in-the-Loop (HITL): Integrated multi-branch feature extractors with downstream models used to recommend the “top-N” health insurance policies to live agents.
Data Science: Developed Graph Convolutional Networks (GCNs) and explainable ML to exploit relational structures in data and provide interpretable predictions, by accessing and processing data via SQL queries, AWS, DataBricks, and distributed computing.

Intern to Full-Time Research Scientist | NASA Ames & SETI Institute | 2015 – 2021

Systems Engineering: Built a neural network from scratch in raw C++ to classify exoplanet spectra for satellite mission planning, demonstrating deep foundational knowledge of machine learning optimization and hardware-level implementation. Deployed a 2D traverse system using LabView and a Cobra probe for wind tunnel surveys. Engineered a web-crawler in Java to post-process aircraft flyover data in noise-mitigation studies. Pioneered preliminary work for my continued research at UoA.

EDUCATION

Ph.D. Computational Science | UCI & San Diego State University | June 2026
M.S. Data Analytics | San Jose State University | 2021
B.S. Applied Physics | University of California, Irvine | 2018

PUBLICATIONS (See Google Scholar for full list)

Johnsen, T. K., et al. “Denoising autoencoder for reconstructing sensor observation data and predicting evapotranspiration: Noisy and missing values repair and uncertainty quantification.” Water Resources Research, 2025.
Johnsen, T. K., et al. “Single and Multi-Mineral Classification using Dual-Band Raman Spectroscopy for Planetary Surface Missions.” American Mineralogist, 2025.
Johnsen, T. K., et al. “SmartDepth: Motion-Aware Depth Prediction with Intelligent Computing for Navigation.” IEEE, DCOSS-IoT, 2025.
Johnsen, T. K., et al. “An Overview of Adaptive Dynamic Deep Neural Networks via Slimmable and Gated Architectures.” IEEE, ICTC, 2024.
Johnsen, T. K., et al. “NaviSlim: Adaptive Context-Aware Navigation and Sensing via Dynamic Slimmable Networks.” IEEE, IoTDI, 2024.
Johnsen, T. K., et al. “NaviSplit: Dynamic Multi-Branch Split DNNs for Efficient Distributed Autonomous Navigation.” IEEE, WoWMoM, 2024.
Johnsen, T. K., et al. “A Multilayer Perceptron for Obtaining Quick Parameter Estimations of Cool Exoplanets from Geometric Albedo Spectra.” Publications of the Astronomical Society of the Pacific, 2020.
Johnsen, T. K., et al. “Elastic Net to Forecast COVID-19 Cases.” IEEE, 3ICT, 2020.

ACCOLADES

Gradslam semi-finalist at UCI (2026).
AGU presenter (2025).
Two-time ACCESS Workshop poster award at SDSU (2024, 2025).
Two year ASSICS grant recipient at SDSU (2021, 2022).
Several guest speaker roles, poster sessions, and formal presentations.