We provide IT Staff Augmentation Services!

Model Predictive Controller Resume

2.00/5 (Submit Your Rating)

TECHNICAL SKILLS:

Top Three Skills: Problem Solving(Advanced), Critical Thinking(Advanced), Coding(Advanced)

Programming Languages: Python, SAS, R, C++, SQL, Spark, Hive, IronPython, JavaScript, XML, WinBUGS

Applicable Software: Spotfire, BayesiaLab, PLINK, KING, Microsoft Office

Big Data Tools: Amazon Web Services

PROFESSIONAL EXPERIENCE:

Model Predictive Controller

Confidential

Responsibilities:

  • Built a Kinematic Model that outputs the steering angle and throttle for a car speeding around a racetrack.
  • Defined the duration of the vehicle model prediction and the interval of time, while considering the latency of the system and computational time.
  • Imposed constraints by understanding the actual limitations of the vehicle design.
  • Designed and tuned the cost functions and their hyperparameters.
  • Trained the Model Predictive Control using the IPOpt optimizer, which passes the current state’s inputs into the optimization solver. The optimization solver outputs the optimized steering angles and throttle for the prediction duration, using the model, cost functions, and constraints set upon the vehicle.

PID Controller

Confidential

Responsibilities:

  • Tuned and built a PID controller that outputs the steering angle for a car driving around a racetrack.
  • The PID controller uses proportions of the current cross track error (CTE), the difference between the most recent CTEs, and the sum of all previous CTEs to obtain the steering angle.
  • Hyperparameters were tuned/optimized using twiddle.

Confidential

Responsibilities:

  • Implemented a 2D particle filter that finds the position of the kidnapped vehicle to within one square centimeter of the vehicle’s location.
  • Initialized particle filter with a map and some noisy GPS estimate localization information of the vehicle’s position, bearing, and velocity.
  • Derived predictions of the vehicle’s location through Gaussian resampling of vehicle’s estimated state.
  • Transformed the new landmark observations into the car’s coordinate system for each sampled prediction and used the Hungarian algorithm to determine the proximity of the landmark observations to actual mapped observations.
  • Based on the proximity, used Bayes rule to obtain probabilities of each sampled predictions and used a robust resampling method to resample from the predictions based on the proximity distribution.
  • Used an average of the resampled predictions to obtain the new estimate of the vehicle’s state.

Confidential

Responsibilities:

  • Implemented an Unscented Kalman Filter (UKF) using the constant turn rate and velocity (CTRV) magnitude motion model to track the position of a pedestrian bicycle with respect to a vehicle.
  • Induced Sensor Fusion to handle both Lidar and Radar measurements fed into the UKF.
  • Tuned the sensor noise and process noise from the measurements and estimates of position.
  • Generated Sigma point values from the current state and augmented the process covariance Sigma points to those Sigma points.
  • Predicted the next state’s Sigma points by passing every generated Sigma point into the process model; used the predicted Sigma points to generate the predicted mean and covariance of the next state.
  • Processed the predicted state into the appointed measurement model based upon the incoming corresponding measurement, obtaining measurement sigma points, mean, and covariance.
  • Updated the state and covariance matrix using the cross - correlation matrix and the Kalman gain matrix.

Confidential

Responsibilities:

  • Implemented the Extended Kalman Filter (EKF) using the constant velocity motion model to track the position of a pedestrian bicycle with respect to a vehicle.
  • Induced Sensor Fusion to handle both Lidar and Radar measurements fed into the EKF.
  • Tuned the sensor noise and process noise from the measurements and estimates of position.
  • Predicted the next state as a function of the current state and its process covariance.
  • Processed the predicted state into the appointed measurement update model based upon the incoming corresponding measurement.
  • Used Jacobian and Hessian matrices create the Kalman Gain matrix and correlation matrix to update the state and its covariance matrix.

Confidential

Responsibilities:

  • Designed a grid searching algorithm that detects and tracks other vehicles near a car’s front-facing camera.
  • Trained a Convolutional Neural Network with similarities of the NVIDIA architecture to discern images of vehicles
  • Slid multiple image patch windows across each incoming image from the car’s camera to detect vehicles
  • Created heatmaps to highlight places likely to possess vehicles on the road.
  • Used a system of deques to help track the position of the vehicles over time. This also reassured the image heatmaps.

Confidential

Responsibilities:

  • Wrote a software pipeline that identifies the lane boundaries in video images from a front-facing camera on a car.
  • Corrected for image distortion from camera lens by measuring the distortion produced from various chessboards taken with the camera.
  • Calibrated camera lens based on the corrections from the chessboards
  • Detected lane lines using masking and thresholding techniques
  • Performed a perspective transform to obtain a bird’s eye view of the lane lines so fitted polynomials could detect the curvature of the lane lines

Confidential

Responsibilities:

  • Trained a vehicle to drive around a racetrack using convolutional neural networks (CNNs).
  • Fired up a simulator to drive the vehicle around the track. Images of the vehicle driving around the track were collected.
  • Used multiple cameras positioned on different places on the vehicle to collect data.
  • Augmented data images to the collection of data. These augmented images focused on overcoming errors, such as being too close to the curb or under/overestimating the steering angle on a turn.

Confidential

Responsibilities:

  • Trained a deep neural network to recognize and classify traffic signs
  • Preprocessed images and augmented samples of image classes based upon the prevalence statistics within the dataset.

Confidential

Responsibilities:

  • Researched and gathered quality features from all US Post-Secondary Institutions.
  • Predicted graduation rate and retention rate using various Machine Learning techniques.
  • Increased transparency of post-secondary institution structures
  • Coded Decision Tree Regressors, Gaussian Mixture Models, and Principal Component Analysis using Python; Microsoft Office

Confidential

Responsibilities:

  • Observed the association between different medications and falling in the elderly population.
  • Results using Bayesian analysis correlate better with results observed in other studies
  • Coded hierarchical one-way normal random-effects model and used Gibbs sampling method in R; Microsoft Office
  • Faced problems with obtaining confidence intervals from large scaled datasets (big data)
  • Invented an algorithm to efficiently bootstrap big data complex sampled datasets
  • Coded the algorithm for the csBLB and ran multiple Monte Carlo simulations to test computational metrics in R; Microsoft Office
  • Assessed the quality of raw exome chip data received from the Confidential
  • Reported discrepancies caused by machine error, human error, or sample contamination
  • Produced final cleaned exome chip data to perform renal study analysis
  • Used PLINK to analyze genotype data; Used KING to analyze family and pedigree checking; Inputted all data and Coded charts and illustrations in SAS; Microsoft Office

Data Scientist

Confidential

Responsibilities:

  • Wrote Spark/Python code that processes raw data from numerous sensors on dozens of oil rigs around the world.
  • Filtered noise and used change point analysis to obtain rig states
  • Classified groups of states as sections to produce key performance indicators that highlight quality of the operations for each rig.
  • Compared and contrasted Machine Learning and Deep Learning results on 2008 PHM data. For machine learning results, I used an ensemble to separate the predictable features and predict the failure of the engines based on statistics from those features. For the deep learning technique, I used a Multi-Scale Convolutional Neural Network.
  • Coded/Used a Multi-Scale Convolutional Neural Network with an additional branch for autoregressive tendencies. Amazon Web Services (AWS) instances were used to train the network.
  • Using AWS, read in las formatted files into a S3 bucket. Using EMR, create a cluster to load the data onto and process the data using a Python/Spark script. Processed datasets are stored in Hive database for future querying.
  • Created interactive visualization demos using Spotfire that used the processed datasets and simulated data to successfully visualize the performance of rig operations and predict performance of future rig operations. Used R to write the data functions and IronPython for specialized script.

Tutor

Confidential

Responsibilities:

  • Helped many students complete projects and pass tests

Teacher

Confidential

Responsibilities:

  • Met second Student Learning Target with over 80% of my small group class passing

Research Assistant

Confidential

Responsibilities:

  • Performed data analysis and quality control in Salt Lake City, UT
  • Researched and studied genetic predisposition in lineage studies
  • Analyzed genotype, phenotype, family, and pedigree data
  • Filtered exome chip data and produced final exome dataset for running analysis

Biostatistician Intern

Confidential

Responsibilities:

  • Used solely R during employment
  • Charged with setting up a parametric survival curve to medical expenditure data
  • Read and researched methods to bootstrap large-scale complex data efficiently
  • Created new method to bootstrap large-scale complex data

We'd love your feedback!