Current focus
PhD candidate (2022–2027)
Hermann-Rietschel-Institut (HRI), TU Berlin
Domain: AI and Energy
Thesis: Building Energy Demand Forecast using Mixture of Experts
AI for a better built environment
Pedram
Babakhani
AI Researcher
Data Scientist
Building Intelligence
I work on machine learning, large language models, smart-building systems, and data-driven energy methods. My research focuses on forecasting, generative models, explainable AI, and intelligent systems for more efficient, comfortable, and human-centric buildings.
Berlin · Germany
People
Spaces
Energy
Data
Intelligence
Spaces
Energy
Data
Intelligence
01
Research
02
Experience
03
Publications
04
Teaching
About me
I am a PhD Candidate in AI and Energy at the Hermann-Rietschel-Institut (HRI), TU Berlin, and I work as a Data Scientist & AI Researcher across HRI and DAI-Labor. My current work sits at the intersection of machine learning, energy systems, LLMs, IoT, and intelligent indoor environments.
Current role
Data Scientist & AI Researcher
DAI-Labor & HRI, TU Berlin
LLMs, ML/DL, recommender systems, interactive spaces, IoT, smart buildings, and energy management.
Research areas
A compact overview of the themes that connect my academic and applied work.
Forecasting & time series
Heat-load forecasting, multi-step prediction, transfer learning, and mixture-of-experts models.
Time seriesTransformersMoEExplainable AI
Interpretability for building-energy models with SHAP-based and attention-based explanations.
XAISHAPInterpretabilityLLMs & NLP
Question generation, subjective reasoning, language-model evaluation, and human-centered AI interfaces.
LLMsNLPIRIoT & smart buildings
Interactive spaces, IoT middleware, smart building infrastructure, and digital assistants in the built environment.
IoTSmart BuildingsEdge SystemsExperience
From FPGA and embedded systems to computer vision, IoT, data science, and AI research.
Oct 2022 — Now
Data Scientist & AI Researcher
Working across AI research and applied intelligent systems with a focus on LLMs, ML/DL, recommender systems, interactive spaces, IoT, smart buildings, and energy management.
- Research on heat-load forecasting, transfer learning, explainable AI, and mixture-of-experts models for building energy systems.
- Development of AI-driven solutions for interactive indoor environments, wayfinding, and care-related digital assistance.
- Teaching, thesis supervision, and team leadership in the IoT & Industry 4.0 context.
- Key project involvement: SPURT, Go-KI, Pflege 4.0, ML-EBESR, and RLT-Auto.
TensorFlowPyTorchLangChainPythonJavaScriptKubernetesDockerSolrGrafana
Apr 2022 — Sep 2022
Data Scientist — IoT
Digital Spine
Berlin, Germany
Worked on IoT solutions for smart buildings and elevators, combining application logic, firmware-adjacent work, and cloud-based workflows.
- Contributed to digital building and elevator infrastructure in an IoT product environment.
- Worked with JavaScript, Python, AWS, Git, and DevOps-oriented workflows.
Nov 2021 — Apr 2022
Data Scientist
Kenkou GmbH
Berlin, Germany
Built data-science components for phone-based photoplethysmography (PPG) and heart-rate variability analysis.
- Developed work around a physiological-signal API using Python, Keras, SciPy, and BioPy.
- Worked with PPG, ECG, and SCG signal-processing and modeling pipelines.
Jan 2021 — Nov 2021
Internship + Master Thesis
u-blox AG
Berlin, Germany
Researched machine-learning-based Bluetooth direction finding for indoor environments.
- Combined TensorFlow, Keras, Flask, Arduino, Raspberry Pi, UWB, and ZigBee tooling.
- Resulted in the IPIN 2021 paper on Bluetooth direction finding using recurrent neural networks.
Dec 2018 — Dec 2021
Hardware Engineer
Trenz Electronic GmbH
Hüllhorst, Germany
Worked in embedded systems and FPGA-related technical documentation.
- Tools included Vivado, C, HLS/SDK, Altium, and Confluence.
- Strengthened the engineering foundation behind later AI and systems work.
Apr 2020 — Dec 2020
Computer Vision & Embedded Developer
PTX Tech
Berlin, Germany
Developed a safety-area determination pipeline using ToF cameras in an embedded/computer-vision setting.
- Worked with Flask, C++, Python, PointCloud, ROS, Raspberry Pi, and Git.
Aug 2018 — Nov 2018
Embedded Software Developer
Assystem AG
Nürnberg, Germany
Worked on remote system update mechanisms for Xilinx 7-series FPGA systems.
- Used VHDL, C, PetaLinux, and Git in the development workflow.
Aug 2017 — Nov 2017
Internship
Bosch Car Multimedia GmbH
Hildesheim, Germany
Implemented LSTM-related work on Zynq UltraScale+ MPSoC platforms.
- Worked with VHDL, Python, and Git.
Projects
Selected projects from recent research and applied development work.
Research collaboration
SPURT
Sustainable Processes of Urban-Rural Transformations. Publicly connected to the research collaboration behind the Opinerium work on subjective question generation.
Applied AI
Go-KI
Project context around practical AI transfer and outreach, including DAI-Labor participation and public presentations.
Care technologies
Pflege 4.0
Competence center for digital solutions in care, including digital support systems for care in the home environment and interactive demonstrators.
Energy AI
ML-EBESR
Development of machine-learning-based control and optimization for efficient and grid-supportive operation of regenerative energy systems.
HVAC systems
RLT-Auto
Data-driven and model-based methods for automatic commissioning and monitoring of air-handling systems under realistic operating conditions.
Public code
Selected GitHub Work
Public repositories that reflect the hardware, machine-learning, and algorithmic side of the portfolio.
Publications
Selected publications and preprints, newest first. You can add or remove items easily by duplicating a publication block below.
2026
Transfer learning and explainable AI for heating load forecasting: A large-scale benchmark with SHAP-based static features
2026
An Explainable Transformer-Based Mixture of Experts for Heat Load Forecasting
2026
Long-term building energy load imputation via exogenous-condition-based generative modelling
2026
A multi-task learning framework for cross-building imputation of long-term missing heating load data
2025
Influence of occupancy data on heating and electricity load forecasting using machine learning
2025
Imputing the long-term missing heating load data using a generative network
2024
GGAS: Ground Guiding Assistant System for Interactive Indoor Wayfinding
2024
Opinerium: Subjective Question Generation Using Large Language Models
2021
Bluetooth Direction Finding Using Recurrent Neural Network
2015
Automatic gamma correction based on the average brightness
Teaching & education
Teaching, supervision, and the academic path behind the portfolio.
Ambient Assisted Living
Main lecturer at TU Berlin.
WiSe 2022SoSe 2023WiSe 2023SoSe 2024Data Science Project
Master course, group supervision at TU Berlin.
SoSe 2023SoSe 2024Multi-Media System
Teaching Assistant at Shahid Beheshti University, Tehran.
SoSe 20152022–2027
PhD Candidate
Energy, Comfort and Health in Buildings
Hermann-Rietschel-Institut, TU Berlin, Berlin, Germany
Thesis: Building Energy Demand Forecast using Mixture of Experts
2017–2021
M.Sc. Data Analytics
Information System and Machine Learning Lab
University of Hildesheim, Hildesheim, Germany
Thesis: Machine Learning-based Bluetooth Direction Finding for Indoor Environments
2011–2016
B.Sc. Computer Engineering
Department of Computer Sciences and Engineering
Shahid Beheshti University, Tehran, Iran
Thesis: ASIC implementation of Automatic Gamma Correction based on Average of Brightness
“Towards intelligent, sustainable and human-centric buildings.”
Get in touch
If you want to collaborate on AI research, energy forecasting, smart-building systems, LLM applications, or applied data-science work, feel free to reach out.
Skills snapshot
Tools & stack
Python, C/C++, JavaScript, Java, TensorFlow, PyTorch, scikit-learn, GPT, Gemini, LLaMA, OpenAI, LangChain, Hugging Face, Flask, FastAPI, Docker, Kubernetes, AWS, GCP, SQLite, MongoDB, Redis, Solr, Arduino, Raspberry Pi, ESP32, ZigBee, BLE, UWB, VHDL, Verilog.