“he is known to me since several years .I know him personally and professionally. he has got great commitment towards work and continuous learning . As of today He has tremendous experience in all areas of AI which includes machine learning ,computer vision ,deep learning using prominent library such as tensor flow, pytorch . In his current job he has been dealing with super challenging problems related to pre and post processing of large size satellite images city area,buildings etcs . I am sure he has got great career ahead and others can take an inspiration from AI journey .”
Palo Alto, California, United States
Contact Info
6K followers
500+ connections
About
Contributions
-
What are the statistical techniques for clustering in ML and how do you choose the right one?
While most of the clustering techniques are easily adoptable by the help of various ML frameworks available, very few ML frameworks are extensible enough to accommodate customer definitions of density, i.e. define what is the meaning of nearness between two samples. While most ML frameworks use Euclidean distance type of measure between various attributes of each sample, but in most of real life data in Enterprise setting is not suitable for RMSE type of Euclidean distance as indication of nearness. So, be sure to provide custom distance metrics for various samples. This becomes much more important for density based techniques like DBSCAN where we have to explicitly provide the value of the epsilon value of the nearness threshold.
-
You're struggling to make sense of your machine learning projects. How can you simplify the process?
Actually the very first step of defining the problem is probably the most important but rarely correctly done step in defining the ML projects. 2 types are issue with defining the ML Problems are: - Overly generalized statement for one wML task - Too Specific or narrow scope of problem for the data at hand. These two above problems are very prevalent in ML projects and we need to achieve the right trade-off between these two extremes by defining the problem statement which is generalized enough while not too specific so the not enough signal to noise ratio is not present in the data at hand.
-
What is the most efficient way to debug Machine Learning code in MATLAB?
One thing which most of Machine Learning engineering fils miss to consider when benchmarking ML models is : "Uncontrolled Random Behavior of different samplers" What it means in simple words is that when you are trying to test ML models, make sure you seed the random generators explicitly with some known seed value. Else any improvement or degrade in performance can be just because of different samples chosen. Hence while testing ML models skipping the seed of random generators sampling is almost a sin!
Activity
-
Shoutout to the team that built https://lnkd.in/g3Y-Zj3W . Really neat site that benchmarks the speed of different LLM API providers to help…
Shoutout to the team that built https://lnkd.in/g3Y-Zj3W . Really neat site that benchmarks the speed of different LLM API providers to help…
Liked by Sandeep Singh
Experience & Education
Licenses & Certifications
-
-
Member
Association for the Advancement of Artificial Intelligence (AAAI)
IssuedCredential ID AAAI Member ID: 637853 -
-
-
-
-
-
-
-
-
-
-
Machine Learning by Stanford University on Coursera
Stanford University
IssuedCredential ID 68KGH4529C2R -
-
-
-
-
-
-
-
-
Volunteer Experience
Publications
-
Building Detection and Location Intelligence in Aerial Satellite Imagery
Deep Learning GT
https://arxiv.org/abs/2302.03156
Building structures detection and information about these buildings in aerial images is an important solution for city planning and management, land use analysis. It can be the centerpiece to answer important questions such as planning evacuation routes in case of an earthquake, flood management, etc. These applications rely on being able to accurately retrieve up-to-date information. Being able to accurately detect buildings in a bounding box centered on…https://arxiv.org/abs/2302.03156
Building structures detection and information about these buildings in aerial images is an important solution for city planning and management, land use analysis. It can be the centerpiece to answer important questions such as planning evacuation routes in case of an earthquake, flood management, etc. These applications rely on being able to accurately retrieve up-to-date information. Being able to accurately detect buildings in a bounding box centered on a specific latitude-longitude value can help greatly. The key challenge is to be able to detect buildings which can be commercial, industrial, hut settlements, or skyscrapers. Once we are able to detect such buildings, our goal will be to cluster and categorize similar types of buildings together. -
Novel Medical Phenotype Approach For Chest X-ray Disease Diagnosis
ML@GT
X‐rays are one of the most common medical tools that doctors and radiologists all over the world have been relying on for decades. With copious X‐ray images and data available, Big Data and Machine Learning techniques can be effectively used for automating diagnosis and providing doctors with an additional helping hand in making correct decisions and reducing error. We will use the publicly available CheXpert chest x‐ray dataset as well as NIH chest X-ray dataset to devise a model that can…
X‐rays are one of the most common medical tools that doctors and radiologists all over the world have been relying on for decades. With copious X‐ray images and data available, Big Data and Machine Learning techniques can be effectively used for automating diagnosis and providing doctors with an additional helping hand in making correct decisions and reducing error. We will use the publicly available CheXpert chest x‐ray dataset as well as NIH chest X-ray dataset to devise a model that can diagnose presence of disease and provide evaluation results. We are going to discuss in detail about data sampling, preprocessing, extraction and training the predictive model. We treat this as a multi‐label classification problem using Convolutional Neural Networks.
Courses
-
MIT Big Data and Social Analytics
-
Projects
-
Grant For Deep Learning Research Fellowship
-
This is Deep Learning Research Fellowship position, where I am responsible for contributing data institute's DL research by contributing to their state of art deep learning framework by employing various DL techniques to various different problem domains.
-
Statistical Bug Isolation Tool
-
Developed a cooperative bug analysis framework using modern statistical predicate analysis
Languages
-
English
Full professional proficiency
Recommendations received
6 people have recommended Sandeep
Join now to viewMore activity by Sandeep
-
Fun times ahead indeed! I'm excited to work with Vincent again! It's a great time to join Waymo!
Fun times ahead indeed! I'm excited to work with Vincent again! It's a great time to join Waymo!
Liked by Sandeep Singh
-
In the fast-paced world of logistics, reducing failed deliveries is crucial for efficiency and customer satisfaction. Here’s how Beans.ai's…
In the fast-paced world of logistics, reducing failed deliveries is crucial for efficiency and customer satisfaction. Here’s how Beans.ai's…
Liked by Sandeep Singh
-
This paper came as crazy as they usually come (in a good way). Hallucination Mitigated by 95%. If everything is true in the paper, it'd open up many…
This paper came as crazy as they usually come (in a good way). Hallucination Mitigated by 95%. If everything is true in the paper, it'd open up many…
Liked by Sandeep Singh
-
My prediction is that in a year's time the hyper-scalers will kill off OpenAI and co. Loved this talk from the Microsoft GenAI team at #CVPR2024…
My prediction is that in a year's time the hyper-scalers will kill off OpenAI and co. Loved this talk from the Microsoft GenAI team at #CVPR2024…
Liked by Sandeep Singh
-
I gave a bit of an unusual keynote talk at ICLR last month. I shared five stories from my 20-year journey in AI so far. It had felt like a bit of a…
I gave a bit of an unusual keynote talk at ICLR last month. I shared five stories from my 20-year journey in AI so far. It had felt like a bit of a…
Liked by Sandeep Singh
-
Absolutely glad to meet Pulkit from Databricks and getting the signed copy of his latest book on databricks. https://lnkd.in/e7vyNEyv
Absolutely glad to meet Pulkit from Databricks and getting the signed copy of his latest book on databricks. https://lnkd.in/e7vyNEyv
Liked by Sandeep Singh
People also viewed
Explore collaborative articles
We’re unlocking community knowledge in a new way. Experts add insights directly into each article, started with the help of AI.
Explore MoreOthers named Sandeep Singh in United States
746 others named Sandeep Singh in United States are on LinkedIn
See others named Sandeep Singh