Ibrahim Ahmed


Education

PhD - Electrical Engineering

Vanderbilt University

Aug '18 - Aug '23

Thesis: Adaptive fault-tolerant control using reinforcement learning.
Courses:
  • Reinforcement Learning

    MS - Electrical Engineering

    Vanderbilt University

    Aug '16 - Aug '18

    Courses:
    • Neural Networks
    • Project in AI
    • Advanced High Performance Computing
    • Computer Vision
    • Embedded Systems
    • Pattern Recognition
    • Solid State Materials
    • Advanced Real-time Systems
    • Advanced Image Processing

    BS - Electrical Engineering and Physics

    Vanderbilt University

    Aug '12 - May '16

    Courses:
    • FPGA Design
    • VLSI Design
    • Algorithms
    • Introductory Applied Machine Learning
    • Modern Physics
    • Modelling and Simulation
    • High Performance Computing
    • Microcontrollers
    • Galactic and Stellar Astrophysics
    • Quantum Mechanics

    Experience

    Software Engineer

    Google

    Apr '25 - Present

    • Lead software infrastructure development for Google’s in-house machine learning accelerators (Tensor Processing Units) to screen for silent data corruption events to strengthen LLM training against hardware faults.
    • Develop statistical models of performance/Total Cost of Ownership informed by silent data corruption events.
    Technology:
    C++
    Python
    SQL

    Senior Controls Optimization Engineer

    Trane Technologies

    Sep '23 - Apr '25

    • Sought out research and development opportunities in AI/ML/Controls software across Trane Technologies' business units and productionized research prototypes.
    • Applied data-driven model predictive control and reinforcement learning for water-side energy optimization of HVACs.
    • Designed LLM Agents for Retrieval-Augmented Generation (RAG), task automation, and analytics pipelines to increase development velocity, reliability, and economy of digital product offerings across multiple business units.
    • Designed AI control algorithms for datacenter cost optimization using digital twins.
    • Democratized data-center cost optimization by designing 3D datacenter layout tools using three.js for simulating energy characteristics under various demand loads.
    • Enabled customers to minimize energy costs of transport refrigeration units by exposing physics-based models to a python API (modelica to FMI) and using Bayesian optimization to automate vehicle configuration.
    • Reduced customers' hardware maintenance costs via predictive maintenance by developing unsupervised anomaly detection algorithms for Variable Refrigerant Flow (VRF) units using steady-state and transient dynamics.
    Technology:
    Python
    Modelica
    Javscript

    Research Associate

    Tennessee State University

    Jun '23 - Sep '23

    • Designed a GUI route planning application using genetic algorithms for autonomous delivery robots in warehouses.
    • Optimized sequence of delivery stops under battery power, congestion, and distance constraints by combining A* search algorithm with custom genetic search operators.
    • Reduced planning latency by 90% down to 1 millisecond by benchmarking routing problems where using brute force would outperform genetic search.
    Technology:
    Matlab

    Graduate Researcher

    Vanderbilt University School of Engineering

    Aug '16 - Jun '23

    • Researched adaptive controller design in a multi-university team for autonomous drone flight under motor faults and wind disturbances using reinforcement learning.
    • Developed embedded Ardupilot flight controller plugins in C++ that work with supervisory reinforcement learning controllers in PyTorch by writing a translation layer for the MAVLink protocol.
    • Won best paper award at AIAA DASC for designing a simulation library for drones in python and C++ using numba and Gazebo.
    • Reduced cooling energy consumption by 2-5% by deploying end-to-end cloud-based machine-learning controllers for energy optimization in 2 campus buildings.
    • Increased sample efficiency on real-world data by training reinforcement learning controllers for smart buildings on data-driven models and physics-based simulations in Modelica and Simulink.
    • Mitigated transfer gap from training to production by writing ETL libraries to parse timeseries building sensor data from third-party vendors and validating simulations against real-world performance.
    • Enhanced building staff’s trust in machine learning controllers by developing a performance logging web application using Dash, with fallback logic for fault-tolerance to infrastructure down-time and data anomalies.
    Technology:
    Python
    PyTorch
    Pandas
    Scikit-learn
    Scipy
    OpenAI gym

    Research Intern

    TieSet Inc

    May '20 - Dec '20

    • As a founding intern, led studies for privacy-preserving experiences in consumer and enterprise electronics.
    • Assured private browsing and non-intrusive ad-serving by designing on-device machine learning library on top of tensorflow.js to match advertisement bids with compatible spaces.
    • Consolidated intellectual property by writing a patent proposal for federated reinforcement-learning for energy optimization in smart buildings without sharing user data.
    Technology:
    Python
    PyTorch
    Flask
    Tensorflow.js

    Applied Scientist Intern

    Amazon Web Services

    May '19 - Aug '19

    • Predicted remaining useful life (RUL) of warehouse machinery by leveraging digital signal processing (DSP) principles (wavelet transforms) for feature-engineering vibration signals for machine learning models.
    • Implemented prognostics research papers into code and conducted feasibility studies by running parallel hyperparameter tuning experiments on m5.24xlarge AWS servers.
    Technology:
    Python
    PyTorch
    Pandas
    Scikit-learn
    Scipy
    OpenAI gym

    Research Intern

    Vanderbilt Department of Physics & Astronomy

    Jun '16 - Aug '16'

    • Validated hypotheses of dark matter distribution in galaxies by developing point-cloud processing algorithms in python.
    • Visualized 3.7 terabytes of n-body simulations as condensed animations by writing a parallel data-mining pipeline executing python scripts using SLURM on campus supercomputer.
    Technology:
    Python (numpy, matplotlib)
    C
    Bash

    Business Intelligence Intern

    Schneider Electric

    Jun '15 - Aug '15

    • Programmed SQL procedures to consolidate 40 million time-overlapping transactions into 1.5 million records with no data loss.
    • Rewrote database backup routines to eliminate redundancies and reduced backup sizes by 75%.
    • Designed visualizations in Domo from Alteryx workflows using sales operations data in North America.
    Technology:
    T-SQL
    Alteryx
    Domo
    Python

    Research Intern

    Institute for Space and Defense Electronics, Vanderbilt University

    Jun '14 - Aug '14

    • Researched effects of solar and cosmic radiation on SRAM memory chips in space.
    • Designed an FPGA interface in VHDL to test memory chips for corruption after irradiation.
    Technology:
    VHDL

    Skills

    Languages

    Python
    C++
    Matlab
    Modelica
    Javascript
    Bash
    Powershell

    Libraries

    OpenCV
    OpenMP
    RTOS
    CUDA
    PyTorch
    Scipy
    Django
    Flask
    Matplotlib
    Plotly
    Dash
    Pandas
    Sklearn
    D3.js
    AngularJS
    jQuery
    SQL
    HTML
    CSS

    Applications

    Ardupilot
    Gazebo
    Simulink
    Dymola

    Extracurriculars

    Graudate Fellow, Vandirbilt Institute for Digital Learning

    Aug '17 - May '18

    • Explored novel uses of digital media (virtual reality, data visualizations) in pedagogy.
    • Scripted and produced videos to distil and convey academic research to a broad lay audience.
    • Curated an online repository of digital learning resources in collaboration with other Fellows.

    Resident Advisor

    Aug '15 - May '16

    • Organized after-hours recreational and educational programs for a residential community of 650 students.
    • Identified students facing academic and personal challenges and provided mentorship, guidance, and connection with on-campus resources.

    Mentor and Technical Specialist - Vanderbilt Design Studio

    Jan '15 - Apr '16

    • Maintained and repaired 3D printers, electronics, and tools at the Design Studio.
    • Assisted users in design process with Makerbot Desktop and Autodesk 123D Design.

    AcFee Subcommittee for Cultural Organizations

    Jan '15 - May '15

    • Reviewed campus organizations’ annual applications for funding.
    • Allocated $218,000 from university funds to 43 organizations based on annual performance analyses.

    Vanderbilt Student Volunteers for Science

    Jan '14 - May '14

    • Taught fundamentals of robotics using LEGO Mindstorms kits to 8th graders in Nashville public schools.
    • Hosted interactive sessions with high school students about careers in engineering.

    Honors

    Vanderbilt 3 Minute Thesis Competition '19

    First Position

    Dean's List

    Fall '12, '14, '15, Spring '13, '14, '15

    Honorable Mention

    International Physics Olympiad '11

    Sigma Pi Sigma

    Society of Physics Students

    Publications

    Ibrahim Ahmed, Marcos Quiñones-Grueiro, Gautam Biswas. 'A high-fidelity simulation test-bed for fault-tolerant octo-rotor control using reinforcement learning' in DASC 2022
    Ibrahim Ahmed, Marcos Quiñones-Grueiro, Gautam Biswas. 'Analysis of the deployment strategies of reinforcement learning controllers for complex dynamic systems' in PHM 2021
    Ibrahim Ahmed, Marcos Quiñones-Grueiro, Gautam Biswas. 'Transfer reinforcement learning for fault-tolerant control by re-using optimal policies' in SysTol 2021.
    Ibrahim Ahmed, Marcos Quiñones-Grueiro, Gautam Biswas. 'Complementary Meta-Reinforcement Learning for Fault-Adaptive Control' in PHM 2020
    Ibrahim Ahmed, Marcos Quiñones-Grueiro, Gautam Biswas. 'Fault-Tolerant Control of Degrading Systems with On-Policy Reinforcement Learning' in IEEE IFAC-PapersOnLine, 2020.
    Avisek Naug, Ibrahim Ahmed, Gautam Biswas. 'Online Energy Management in Commercial Buildings using Deep Reinforcement Learning' in IEEE International Conference on Smart Computing, (SMARTCOMP), 2019.
    Ibrahim Ahmed, Gautam Biswas, Hamed Khorasghani. 'Comparison of model predictive and reinforcement learning methods for fault tolerant control' in IFAC-PapersOnLine, Warsaw, Poland, 2018.