I’m an experienced Manager of Software and Data Science (Lead 15+ Developers Simultaneously)
- โ ๐ ๐ฎ๐บ ๐ฎ๐ป ๐ฎ๐๐๐ผ๐ฑ๐ถ๐ฑ๐ฎ๐ฐ๐, ๐๐ต๐ฎ๐ ๐ฐ๐ฎ๐ป ๐๐ฎ๐ธ๐ฒ ๐ฎ ๐ฝ๐ฟ๐ผ๐ท๐ฒ๐ฐ๐ ๐ณ๐ฟ๐ผ๐บ ๐ถ๐ป๐ฐ๐ฒ๐ฝ๐๐ถ๐ผ๐ป ๐๐ผ ๐ฑ๐ฒ๐น๐ถ๐๐ฒ๐ฟ๐, ๐ฎ๐ป๐ฎ๐น๐๐๐ถ๐ป๐ด, ๐ถ๐ฑ๐ฒ๐ป๐๐ถ๐ณ๐๐ถ๐ป๐ด ๐ฎ๐ป๐ฑ ๐ฝ๐ฟ๐ผ๐๐ถ๐ฑ๐ถ๐ป๐ด ๐๐๐ฟ๐๐ฐ๐๐๐ฟ๐ฒ๐ฑ ๐ฎ๐ป๐ฑ ๐ฎ๐๐๐ผ๐บ๐ฎ๐๐ฒ๐ฑ ๐ฎ๐น๐ด๐ผ๐ฟ๐ถ๐๐ต๐บ๐ถ๐ฐ ๐๐ผ๐น๐๐๐ถ๐ผ๐ป๐ ๐ณ๐ผ๐ฟ ๐ฏ๐๐๐ถ๐ป๐ฒ๐๐ ๐๐๐ฎ๐ฟ๐๐ฒ๐ด๐ ๐ป๐ฒ๐ฒ๐ฑ๐ ๐๐ผ ๐ฒ๐ป๐ต๐ฎ๐ป๐ฐ๐ฒ ๐ฐ๐ผ๐บ๐ฝ๐ฎ๐ป๐ ๐ด๐ฟ๐ผ๐๐๐ต
โ ๐ ๐ฐ๐ผ๐ป๐๐ถ๐ฑ๐ฒ๐ฟ ๐บ๐๐๐ฒ๐น๐ณ ๐ฎ ๐๐๐ฟ๐ผ๐ป๐ด ๐ฐ๐ผ๐บ๐บ๐๐ป๐ถ๐ฐ๐ฎ๐๐ผ๐ฟ, ๐ถ๐ป๐ณ๐น๐๐ฒ๐ป๐ฐ๐ฒ๐ฟ ๐ฎ๐ป๐ฑ ๐ฐ๐ฟ๐ผ๐๐ ๐ฑ๐ฒ๐ฝ๐ฎ๐ฟ๐๐บ๐ฒ๐ป๐ ๐ฐ๐ผ๐น๐น๐ฎ๐ฏ๐ผ๐ฟ๐ฎ๐๐ผ๐ฟ
โ ๐ ๐ฎ๐บ ๐ฎ๐ป ๐ฒ๐ ๐ฝ๐ฒ๐ฟ๐ถ๐ฒ๐ป๐ฐ๐ฒ๐ฑ ๐ฃ๐ฟ๐ผ๐๐ผ๐ฐ๐ผ๐น ๐ฅ๐ฒ๐ด๐๐น๐ฎ๐๐ผ๐ฟ, ๐ ๐ต๐ฎ๐๐ฒ ๐๐ฒ๐ ๐๐๐ฐ๐ฐ๐ฒ๐๐๐ณ๐๐น ๐ฝ๐ฟ๐ผ๐๐ผ๐ฐ๐ผ๐น๐ ๐ฎ๐ป๐ฑ ๐ฟ๐๐น๐ฒ๐ ๐ณ๐ผ๐ฟ ๐ฒ๐ณ๐ณ๐ถ๐ฐ๐ถ๐ฒ๐ป๐ ๐ฐ๐ผ๐บ๐ฝ๐ฎ๐ป๐ ๐๐ผ๐ฟ๐ธ
Iโm proficient in the following sciences: Mathematics,ย Physics,ย Computer Science,ย Statistics
Mathematics:
Final Israel Open University Grades:
Statistics: 96 out of 100
Discrete Math: 96 out of 100
Differential And Integrable Math: 82 out of 100
Linear Algebra: 91 out of 100
- Proof by induction, Proof by contradiction.
-
Linear Algebra – Vector-Matrix Linear Transformation,Pre-image of sets,subspaces,subsets, ranges under transformations, Image of Transformation,Rotations,Scaling, R2,R3,Rn, Unit Vectors, Projections, Distributive Properties, Associativity Properties ,Composition Properties, Inverse Determinant Properties,Transpose, Null space, Column space, Rank,Reduced Row-Echelon Form ,Linear Independence,Matrix Powers via Diagonlization.
- Concept of Set, Injection, Surjection and Bijection, Concept of groups, Analysis, , Differential equations, Sequences, Series, Convergence, Limits, Calculus , Continuity, Ordinary differential equations (ODE), partial differential equations (PDE) and stochastic differential equations (SDE).
-
Orthogonal Complements, Projection Matrix,Least squares approximation,basis change,Orthonormal Bases using the Gram-Schmidt Process, EigenDecomposition (Eigenvalues and Eigenvectors),SVD, Covariance Matrices and PCA Reduction, Vector/Matrix dot and cross products.
- Euclidean Geometry ,Non-Euclidean Geometry, Proofs,ย Projective Geometry, Sphere Geometry, Hyperbolic Geometry,Trigonometry,Fourier Analysis, Signals Analysis, Time Series Analysis.
Physics:
- Displacement ,Acceleration, Velocity ,Kinematic Formulas, Projectile Motion
- Centripetal Force, Centripetal Acceleration, Gravity,Mass,Weight
- Work, Potential Energy, Kinetic Energy, Thermal Energy, Power, Springs, Hooke’s Law, Conservation of Energy, Mechanical Advantages.
- Angular Velocity and Speed,Rotational Kinematic Formulas,Moments,Torque Moment of Inertia,Rotational Inertia,Rotational Kinetic Energy,Angular Momentum,Cross Product and Torque
- Normal Force, Contact Force,Balanced and Unbalanced Forces,Force of Friction ,Static and Kinetic Friction, Force of Tension
Computer Science:
- Operating Systems, Data Structures ,Set theory,Combinatorics,Algorithms,Computer Graphics,Logic,Image Processing and Analysis,cyber security, Signal, Image And Data processing,Computer Architecture,Random graphs,Complexity Theory, Storage Systems, Internet of Things (IOT),Internet Networking,Computability, Software Verification, Distributed Systems,Compilation,Database Management Systems,Reverse Engineering,Artificial Intelligence,Cryptology,Robotics,Concurrent Programming,Computer Vision ,NLP, Reinforcement Learning,Computational Geometry,Data Analysis, File Compression
Statistics:
- Statistical Methods,Measures of location,Central Tendency,Dispersion,Probability Theory,Random Variables
Expectation of Random Variable,Standard and Continuousย probability distributions,
Bivariate and multivariate Distributions,Bivariate Transformations. - Correlation and Regression,Limit Laws,Order Statistics, Sampling Distribution,Tests of Significance
Index Numbers,Time Series.
ย
I have proven work experience as a R&D Engineer in: Data Scientist, Computer Vision, Autonomous, Algorithm, Python
Data Scientist:
- Deep Machine Learning
- Designer/Trainer/Researcher.
- Fully Connected Networks , Convolutional Networks
- Train/Validation/Inference Increasing accuracy:
- ย Over-fitting: Regularization(L1 sum of Absolute Coefficients for sparse, L2 sum of squared for dense,lambda for taming), Dropout, Batch Normalization,Early-stop.
- General Hyper Parameter Tuning: Learning rates, Optimizers, Grid-Search.
- Under-fitting: Unbalanced Data, Data Augmentation, Bias/Variance Trade-off Tuning.
- Data Preparation -Cleansing,Imputing vs Generalization Tradeoff, Data Preprocessing -Dropping/Dummy Vars/Categorical-Discrete Continuous Encoding, Tensoring.
- False Positive(T1)/False Negative(T2) Errors(Wrongs), Accuracy (TP+TN/Total),Recall (TP/TP + FNย ย –ย ย T2 penalized), Precision(TP/ TP+FP –ย ย T1 penalized)
- Harmonic Mean(F1) – 2โ PrecisionโRecallโ/Precision+Recall, F-beta (1 + \beta^2)โPrecisionโRecallโ/beta^2โPrecision+Recall, Roc Area(TP/TotalT,FP/TotalN)
- Mean Absolute,Mean Squared(Differential),R2 (MSE for predicted model/MSE for average line)
- Supervised:Classification, Linear& Polynomial&Logistic Regression, SVM(kernel trick, c-value(lower margin better classification),RBF Kernel-mountain gammas), K-Dimensional Tree,Decision tree(information gain/entropy states reduction by approximation on sample batches for fast convergence/(Shannon entropyย by sum of probability logs of difference of parent and average of it’s children instead of tiny probability products)).
- Unsupervised Clustering:ย KNN, K-means(scree plot elbow) . Gaussian Mixture Models. Cluster Validation,anomalities,Hierarchical,ย Density
- Unsupervised Dimentionality Reduction: Random Projection , Independent Component Analysis ,PCA
- RNN, Attention (multiplicative and additive), LSTM, GRU.
- Transfer Learning.
- Up-Transpose (DeConvolutions)/Aliasing.
- Perception and Detection based deep learning, semantic segmentation(YOLO(Iou/Nms), RCNN, instance segmentation).
- VggNet, ResNet, GoogLeNet/Inception ,MobileNet, HF-Net.
- NLP – Captioning NN’s, Feature Extraction,TF-IDF,Topic modeling,seq2seq,ASR,CFG
- Behavioral Cloning.
- Adverserial Training(GAN).
- Deep Reinforcement Learning(Deep Q-network,Actor-Critic).
Computer Vision:
- Detection
- Feature Extraction – SIFT(scale robust), SURF and HOG(noise robust),ย Hough Transforms(polar spaces parameterization as lines xcoos-ycos = d),Generalized Hough Transforms(displacement vectors r,ย gradients theta, peaks ), Color Spaces(e.g. keep yellow by converting in HSV instead of loosing it in grayscale) , Spatial Information, Color Histograms.
- Laplacian (second derivative filter), Difference of Gaussians , Canny Edge Detection (high threshold joined by lower threshold)
- Sliding windows/patches with overlap, changing small scales for horizons and large for closer images.
- Ray Tracing with floats, with integers(bresemham).
- Template matching with threshold.
- False-Posivite reduction with next frame threshold,
- Duplicate reduction using Iou Centers and Nms.
- Tracking(moving predictions with detection as a closed cycle loop)
- Line detection applying polynomial calculations of curves and estimation of curvature through radius by applying 2nd derivatives.
- 3D Perception
-
- 3D point clouds.
- “Voxels” using PCL.
- Bundle Adjustments.
- Triangulation.
- Eigen Vector Decomposition for high dimensional domain computations in different basis spaces.
-
- General
-
- Recognition and Classification by Prior Generatives and by Posterior Discriminatives in Detection/Verification/Identification Instances and Categorization/Scene/Context Generics.
- Computing the radial/tangential distortion rectifying transformation matrix.
- Perspective transforms -fitting polynomials, masking , thresholding,ย parallel line convergence fitting ,smaller to bigger enhancement, map(top view/bird view).
- ย Thresholding angles (such as gradients) using trigonometric functions such as Arctan2.
- Camera matrix transformation from 3d to 2d.
- Interpolation using linspace.
-
Autonomous:
- Estimation
- Sensors – Lidars(Infra red (velodyne) rain or fog – cycles pulses of point clouds), Radars(radio wave doppler, adaptive cruise control,velocity measurement rain/fog/underground fov),Accelerometers, Gyros.
- Extended Kalman Filters(unimodal), Unscented Kalman Filters, Particle Filters(multimodal),Monte-Carlo(multimodal).
- Multi Sensor Fusion – DeMorgansย law for independent measurements followed by max
- RANSAC, PCA/T-SNE.
- Gaussian Naive Bayes Probabilities (estimating measurements and noise), Maximum Likelihood, Continuous Probabilities.
- Pedestrian location heading and speed.
- Localization
- rotations/quaternions(avoiding gymbel lock)/Euler (visualization).
- Image-based.
- Sensor Fusion Based.
- Simultaneous Localization and Mapping (SLAM).
- IMU for localization.
- GNSS+RTK (real-time kinematic) – High accuracy of GPS.
- Mapping
-
- Configuration space.
- Dense mapping vs. Sparse mapping(landmarks).
- Structure from Motion(SFM).
-
- Motion Path Planning(Map,Goal Value,Cost Value)
- Trajectory over a fixed receding time horizon.
- A* search and navigation algorithm around obstacles,Dynamic A*.
- Partially observable Markov decision processes (POMDPs).
- Behavior tree/Finite state machine motion.
- Detour,Lane shift Policies.
- Optimal min neighbor policy.
- Probabilistic roadmap -Dijkstra’s shortest path , Voronoi graph and medial axis graphs.
- Two-and-a-half-dimensional , Random Sampling,
- Rapidly-exploring Random Tree (RRT).
- Dynamic programming – for each point there is a path(not just the start).
- Control
- PID Control – cross track error. (proportional , integral, differential )
- Model Predictive Control (MPC)- optimize control actuations over a given time horizon.
- Bicycle model(steering angle, brake, heading)
- Ackerman steering model.
- Differential drive model for robots.
- Slip angles and slip ratios.
- Linear-quadratic regulator (LQR).
- Twiddling techniques.
- Statics(equilibrium),Kinematics(forces),Kineto-statics(1 time frame),Dynamics(all time frame) models.
Algorithm:
-
- Sorting and searching
- Graph algorithms:
- Graph traversal (DFS, BFS) and applications
- Minimum spanning tree
- Shortest path
- Matchings
- Network flow
- Traveling salesman problem
- Hamilton cycle
- Algorithmย design:
- Divide-and-conquer
- Graph traversal
- Greedy
- Dynamic Programming(store the results of subproblems)
- Reductions
- Use of advanced data structures
- Algorithmย correctness:
- Proofs and proof techniques (assumptions, basic logic inference and induction)
- Algorithmย analysis:
- Time and space complexity
- Asymptotic analysis Big O(Theta,Omega,O)
- Worst/Best/Average case analysis
Python:
- NN Frameworks – Pytorch,Tensorflow,Keras,Lua,KerasRL, Tensorforce
- DataScience -(Python, Pandas, NumPy, SK-Learn, Shapely)
- Autonomous – Networkx, PyMC
- Computer Vision – PIL, OpenCV, SK-Image,PCL
- Visualization – Matplotlib,Plotly
- NLP -Nltk, Textblob, Spacy
โ Further Programming Languages I have work experience with(to state only a very few):
-
- c++ – eigen, boost (Basic)
- c# – WPF,WCF,Web Api (Expert)
- Ecma javascript 2019 ,React, Angular 5,TS (Moderate)
- Matlab(Basic)
โ Professional .NET developer: Web/ Desktop Apps; Multi-Thread Async; P/Invoke; Design Patterns; Architecting/Framework designer;
โ Experienced system analyst: Business needs delivery focused. Proficient in identifying and setting priorities.
โ Fast Learner and Problem solver: Highly motivated, can contribute to any project. Productive in both team-based and self-managed projects, including system design and technical documentation, peer training and review. Excellent English; written and verbal