Hi, I'm Chandan Dwivedi
Senior Machine Learning Engineer
Passionate about Exploring and builduing AI solutions that can transform lives!
Contact MeAbout Me
My Introduction
Passionate Machine Learning Engineer with solid Computer Vision, Artificial Intelligence and Machine Learning development background I have good experience in building scalable AI solutions using latest technologies like MLOps and microservices. I have huge interest in Edge AI and have experience in writing efficient low-level C/C++ codes and optimizing AI models for Edge devices.
Skills
My Technical LevelDevelopment
All About the CorePython
90%C++
90%R
80%PySpark
75%Scala
60%SQL
75%MS Excel
70%JAX
30%Frameworks
Everyone Needs SupportNumPy
90%pandas
90%matplotlib
80%scikit-learn
85%Spark MLlib
70%Pytorch
85%TensorFlow
85%OpenCV
90%Pillow
85%NLTK
50%ONNX
80%seaborn
70%Flask
85%FastAPI
80%Machine Learning
Theory, theory!Linear and Logistic Regression
95%Decision Trees
95%Ensemble Models
90%Clustering
65%Convolutional Neural Networks
95%Graph Neural Networks
50%Recommender Systems
75%Natural Language Processing
65%Exploratory Data Analysis
90%Multi-modal Learning
70%Time Series
55%GANs
65%Cloud and Engineering
Fly Fast & High!Kubernetes
75%Docker
80%AWS Athena
65%AWS Inferentia
60%AWS Lambda
70%Apache Flink
40%Kafka
60%Weights & Biases
80%Databases and Viz
Wow! FactorMySQL
75%MongoDB
85%PostgresSQL
75%Tableau
70%Power BI
40%Plotly
70%Qualification
My Personal JourneySenior Machine Learning Engineer
Zupee
-
Improved the performance of the fraud detection solution using Deep Learning and Neural Networks to detect similarity in profile pictures
-
Part of the team builduing Real-time fraud detection system architecture. Worked on exploring various solutions and proposing development flow
Senior Software Engineer
FYND
-
Lead the AI team and Implemented media content moderation, recommendation engine, Text moderation using NLP, 3D avatar generation using selfie images, and various other AI features for the E-gaming platform.”
-
Worked on the complete lifecycle of AI products right from data collection, data cleaning, Model training, tuning, and optimizing to deploying models on AWS Kubernetes pods and/or mobile devices
-
Increased the performance of the Augment Reality (AR) virtual product trail platform by 25% by using NCNN and Multithreading along with Web Assembly. The core APIs are written in C++ language.
-
Increased the performance of the Augment Reality (AR) virtual product trail platform by 25% by using NCNN and Multithreading along with Web Assembly. The core APIs are written in C++ language.
-
Leveraged Knowledge in C++, Computer Vision, Deep Learning, NLP, Recommendation system, Pytorch, TensorFlow, TF Lite, OpenCV, NCNN, WebAssembly, Emscripten, Machine Learning and SDLC
Machine Learning Engineer
DeepEdge.ai
-
Worked on EDGE-AI. Implemented Face Recognition pipeline on a camera device which uses Ambrella's CV2 Computer Vision "System-on-Chip"(SOC). This device is being used by security forces of various countries.
-
Performed research on architectures, trained models, built end-to-end pipelines in both python & C++ and finally deployed models on SOC after performing optimizations like Quantization and prunning. Compiled the model with SOC's SDK inoder to deployt it.
-
Interacted with various clients and proposed and created various POCs. Achived very good feedbacks from clients.
-
mplemented efficient, fault-tolerant C++ libraries (.so and .dlls) for Face Registration Engine and used .NET application to call and use the libraries
-
Performed the testing of AI products and benchmarked the deployed models on Edge devices using real world data.
Software Development Engineer
Aiseon Healthcare Technologies Pvt. Ltd.
-
Developed a scalable, distributed and asynchronous Rest API for Cognitive models by using Message Broker, Task queue, Gunicorn, Nginx, Docker and Kubernetes. Made it scalable and robust to be able to serve a high number of concurrent demands and integrated Continuous Integration and Continuous delivery (CI/CD) and auditing (MongoDB)
-
Increased performance of AI platform by 20% by using hybrid and decoupled Deep Neural Network models.
-
Redesigned the Job orchestration and pipeline to process data, train networks and generate results by using IBM-Cloud Bare Metal Servers, Spark MLlib, IBM-Cloud Python runtime, TensorFlow, and OpenCV
-
Developed new features for data annotation tools to make annotation easy and more automatic using Neural networks, clustering algorithms, segmentation models, multithreading, and multiprocessing
-
Developed new features for data annotation tools to make annotation easy and more automatic using Neural networks, clustering algorithms, segmentation models, multithreading, and multiprocessing
-
mplemented A/B Testing using Kubernetes canary deployments
B.Tech in Computer Science and Engineering
Motilal Nehru National Institute of Technology, Prayagraj, IndiaHigher Secondary in Science
Little Flower House, Varanasi, IndiaSecondary
Jawahar Navodaya Vidyalaya, Varanasi, IndiaPortfolio
My Projects
DR leasions detection on retinal images
Computer Vision and Deep Learning
Detection and segementation of 8 different lesions of Diabetic retinopathy(DR) which causes serious complications in eye.
Implemented and trained various deep learning models to detect and segment different lesions which inturn helped identifying severity if desease in patients.
Experimented with various architectures like U-Net, Mask R-CNN, Faster-RCNN with Inception V2(Tensorflow-api) etc.
Tech Stack
Research Papers Referred
View Code
Game Recommendation System from Gameplay Data
Recommendation system and Machine Learning
Built a Game recommendation system leveraging user gameplay history data and user social data.
Tuned hyperparameters for three CF approaches: KNN, SVD and MF to obtain good results
Experimented with neural MF and obtained considerable results.
Tech Stack
View Code
Face Recognition
Convolutional Neural Networks
Implemented face recognition system using state of the ART architecture: Arcface and Retinaface.
Focused on parameter tuning and optimization techniques to achive realtime inference and high accuracy on even distant and small faces.
Explored various datasets like MS1M, Widerfaces, DeepGlint's dataset etc.
Implemented tracking mechanisms by using tarcking algorithms like SORT, Deep-Sort, Centriod tracking etc.
Tech Stack
Research Papers Referred
ArcFace: Additive Angular Margin Loss for Deep Face Recognition
RetinaFace: Single-stage Dense Face Localisation in the Wild
Simple Online and Realtime Tracking with a Deep Association Metric
Image Background Remover
Convolutional Neural Networks
Built Image background remover application using Convolutional NN, and exposed it as a REST API. Deployed the API using microservices architecture.
Tuned the parameters of the model to make model work better around edges.
Tech Stack
Avatar Generation from selfie images
Deep Learning
Researched on GAN model and autencoder based models to generate 2D and 3D avatars from selfie images
Explored low latancy models for 2D avatar generation
Worked on implementing pipeline for avatar generation
Tech Stack
Content Moderation for Images
Convolutional Neural Networks
Developed and deployed a neural netowrk model to moderate images for explicit, NSFW, and Gore content
Downloaded ~500k images for the dataset from various sources. Performed data cleaning and created dataset for training and Benchmarking the model
Implemented SOTA model EfficientNet-V2 and trained the model on the above dataset. Used various optimization methods like One-cycle policy, sampling, regularization etc. to improve the performance of the model.
Tech Stack
Research Papers Referred
EfficientNetV2: Smaller Models and Faster Training
Xception: Deep Learning With Depthwise Separable Convolutions
Super-Convergence: Very Fast Training of Neural Networks Using Large Learning Rates
Undergrad Final Year Project: Self Driving Car
CNN, Computer Vision and Edge AI
Mounted Raspberry pi on RC car and used Raspberry Pi camera to make the RC car a Self Driving Car
Trained and deployed models to detect humans and objects, detect and classify traffic signs, detect lanes etc. and deployed on raspberry pi.
Worked on path planning of Self Driving car in a udacity's SD-pathfinding Simulator to test our path planning algorithm.
Tech Stack
Research
My PublicationsTBD
TBD
Abstract
I am Researching on GANs. It is expected to get completed soon...
TBD
TBD
Abstract
I am also Researching and exploring Graph Neural Networks. It is expected to be completed soon...
Certifications
Extra Courses I have Undertaken
Deep Learning Specialization
Expiry Date: Does not expire
Sequence Models
Expiry Date: Does not expire
Improving Deep Neural Networks: Hyperparameter Tuning, Regularization and Optimization
Expiry Date: Does not expire
Convolutional Neural Networks
Expiry Date: Does not expire
