“Devesh is a powerhouse of Data Science! I had the pleasure to work closely with Devesh for 3 years at Unilever, where he made a dramatic positive step-change to UL's Data Science capability, both by his own outputs and the positive influence his leadership and behaviour had on those around him. Devesh is super curios. He is a passionate learner and learning appears to come easy to him. For someone who is so hands on with his data science Devesh also has a natural ability to translate the technical to non-technical stakeholders and appears un-phased where others might feel daunted. I very much enjoyed and valued my time working with my good friend Devesh and plan to remain in contact ongoing - whether this be to talk naturally about language processing; randomly about forests; or testingly(?) about cricket!”
About
Devesh is a seasoned leader (17+ years) in the field of AI & Machine Learning- leading a…
Contributions
-
What do you do if you're facing conflicts in data science and need to find a resolution?
Understand Context: Identify data sources and variables causing discrepancies. Types of Conflicts: Semantic, structural, and statistical. Data Profiling: Examine distributions and outliers. Data Cleaning: Standardize, impute missing values, and handle outliers. Conflict Resolution: Manual inspection, domain knowledge, consensus, data fusion, and ML models. Validation and Testing: Metrics and unit tests. Iterate: Continuously refine your approach.
-
How do you assess learner characteristics such as prior knowledge, goals, and preferences?
The process of assessing prior knowledge involves several key steps: Expert Knowledge:Engage with domain experts to gather qualitative insights and translate them into quantitative priors. Hierarchical Modeling:Use hierarchical structures to incorporate multiple levels of prior knowledge. E.g.-global priors can be set for a group of parameters, while allowing individual parameters to vary. Conjugates: Choose priors that are conjugate to the likelihood function, which simplifies the computation of the posterior distribution. Noninformative Priors: When little prior knowledge is available, noninformative priors, which have minimal impact on the posterior distribution, can be used. These priors are designed to be as neutral as possible.
-
How do you deal with vanishing or exploding gradients in CNN backpropagation?
Gradients play a fundamental role in the field of machine learning, particularly in the optimization of algorithms. In essence, a gradient is a vector that points in the direction of the steepest increase of a function. In the context of machine learning, this function is typically a loss function that measures the difference between the algorithm’s predictions and the actual data. Gradients are crucial because they guide the optimization process; by following the negative gradient, an algorithm can adjust its parameters to minimize the loss function. This process, known as gradient descent, is central to training models such as neural networks. The gradient tells us how to change the parameters to make the model more accurate.
-
What do you do if your AI career decisions are influenced by emotional intelligence?
EI is a pivotal component in the realm of AI careers, it provides the necessary human touch that ensures AI technologies are developed with empathy and understanding. Professionals with high EI are adept at navigating the complex social dynamics of the workplace, fostering collaboration, and driving innovation. They are also better equipped to design AI systems that are attuned to human emotions. Moreover, EI helps AI practitioners to understand and address the ethical implications of their work, ensuring that AI solutions are fair, unbiased, and beneficial to society. Thus, EI is not just a complementary skill but a core competency that enhances the effectiveness and humanity of AI solutions.
Activity
-
Meet our eminent 3AI Thought Leader. Divik Saxena, Vice President & Chapter LeaderAnalytics & Research, Fidelity Investments 3AI’s 1100+ AI &…
Meet our eminent 3AI Thought Leader. Divik Saxena, Vice President & Chapter LeaderAnalytics & Research, Fidelity Investments 3AI’s 1100+ AI &…
Liked by Devesh Raj
-
Looking to fine-tune GPT4o? Why not try the open source alternative with Phi 3.5 distilled from GPT4? Finetuning Phi 3.5 is 2x faster and uses 50%…
Looking to fine-tune GPT4o? Why not try the open source alternative with Phi 3.5 distilled from GPT4? Finetuning Phi 3.5 is 2x faster and uses 50%…
Liked by Devesh Raj
-
#FidelityInvestmentsIndia recently organized a #CommunityRoadshow2024 across our Bangalore and Chennai offices. The roadshow was aimed at promoting…
#FidelityInvestmentsIndia recently organized a #CommunityRoadshow2024 across our Bangalore and Chennai offices. The roadshow was aimed at promoting…
Liked by Devesh Raj
Experience
Education
-
Indian Institute of Technology, Hyderabad
The three year course covered coursework in the subjects of Probabilistic ML, Applied Machine Learning, Deep Learning, Statistics and Research work- Developing a novel framework for feature representation and feature reinforcement on structured as well as unstructured data.
-
-
63rd rank in State entrance exam
-
-
Licenses & Certifications
Publications
-
1) Responsible AI is good conscience and good for business too
Analytics Insights magazine
Importance of AI Ethics in developing AI systems
-
2) Enhancing Navigation through an Intelligent Computer Vision based System- IEEE -International Conference on Convergence of Technology (I2CT)
IEEE-International Conference on Convergence of Technology (I2CT).
-
3) Decision System Model For Predicting Cardiovascular Disease
IISC-IIMB Data Science International conference
-
4) Feature Reinforcement using Autoencoder -Asian Journal of Convergence in Technology (AJCT)
Asian Journal of Convergence in Technology (AJCT).
Patents
-
A Framework to predict System Behavior through Automated Hierarchical Network
GB 20208333.3
The invention relates to methods and apparatus for predicting system variables using a Bayesian network. It then enables simulation and optimization of the variables.
-
System and Method for Grading of roads using Computer Vision
DE 201811048223
The health of the road can be quantified as a function of -grade of potholes, proximity of potholes, luminosity, width of the roads. This number can then be ported to maps so that the driver can make more informed decisions.
Courses
-
Bayesian Machine Learning
-
-
Computer Vision
-
-
Deep Learning
-
-
Machine Learning
-
-
Natural Language Processing
-
-
Neural Networks
-
-
Probabilistic Machine Learning
-
-
Scala
-
Honors & Awards
-
Maths Olympiad
-
Rank 7th
-
Semi-Finalist
UNICOM-Analytics & Big Data Olympiad'2017
500+ companies participated in the Analytics & Big Data Olympiad held by UNICOM.
-
Speaker at Artificial Intelligence Connect(AIC)
-
Deep Learning
-
Speaker at Artificial Intelligence World Congress, London
-
Deep Learning model customization for domains
-
Speaker at NASSCOM summit
-
Deep Learning
Test Scores
-
XAT 2007
Score: 98.6 percentile
XAT 2007-XLRI- scored 98.6 percentile
-
State Engineering entrance exam
Score: 585
63rd rank in State engineering entrance exam
Recommendations received
3 people have recommended Devesh
Join now to viewMore activity by Devesh
-
Humbled to join 3AI Thought Leaders’ group! Looking to learn and share from some of the most eminent leaders in analytics space!
Humbled to join 3AI Thought Leaders’ group! Looking to learn and share from some of the most eminent leaders in analytics space!
Liked by Devesh Raj
-
Humbled to join 3AI Thought Leaders’ group! Looking to learn and share from some of the most eminent leaders in analytics space!
Humbled to join 3AI Thought Leaders’ group! Looking to learn and share from some of the most eminent leaders in analytics space!
Shared by Devesh Raj
-
Meet our eminent 3AI Thought Leader. Devesh Raj, Director - Data Science & AI CoE, Fidelity Investments 3AI’s 1100+ AI & Analytics Thought Leaders,…
Meet our eminent 3AI Thought Leader. Devesh Raj, Director - Data Science & AI CoE, Fidelity Investments 3AI’s 1100+ AI & Analytics Thought Leaders,…
Liked by Devesh Raj
-
Shankar Swaminathan, VP – Analytics, Research and Data at #FidelityInvestmentsIndia, is a panelist at the nasscom Global Inclusion Summit 2024 on…
Shankar Swaminathan, VP – Analytics, Research and Data at #FidelityInvestmentsIndia, is a panelist at the nasscom Global Inclusion Summit 2024 on…
Liked by Devesh Raj
-
🎉 Huge congratulations to Fidelity Investments, winners of the 3AI Standout AI & Analytics Solutions Awards 2024 !! Kudos to all the dedicated teams…
🎉 Huge congratulations to Fidelity Investments, winners of the 3AI Standout AI & Analytics Solutions Awards 2024 !! Kudos to all the dedicated teams…
Liked by Devesh Raj
-
After the exciting releases of relative "large" large language models last week (Llama 3.1 405B and Mistral Large v2), it's refreshing to see Google…
After the exciting releases of relative "large" large language models last week (Llama 3.1 405B and Mistral Large v2), it's refreshing to see Google…
Liked by Devesh Raj
-
A very insightful read by Chip Huyen on building generative AI platform. https://lnkd.in/gKGHn5N9
A very insightful read by Chip Huyen on building generative AI platform. https://lnkd.in/gKGHn5N9
Shared by Devesh Raj
-
Last week, two news caught my attention. First one was the launch of SearchGPT, a cutting-edge search GPT from OpenAI. As SearchGPT steps into the…
Last week, two news caught my attention. First one was the launch of SearchGPT, a cutting-edge search GPT from OpenAI. As SearchGPT steps into the…
Liked by Devesh Raj
-
The nasscom 𝗥𝗲𝘀𝗽𝗼𝗻𝘀𝗶𝗯𝗹𝗲 𝗔𝗜 𝗛𝘂𝗯 recently held a stakeholder consultation to refine and improve the practical utility of its upcoming…
The nasscom 𝗥𝗲𝘀𝗽𝗼𝗻𝘀𝗶𝗯𝗹𝗲 𝗔𝗜 𝗛𝘂𝗯 recently held a stakeholder consultation to refine and improve the practical utility of its upcoming…
Liked by Devesh Raj
-
Lyft used graph embeddings to improve riding experience for 20M+ riders 🚙 — by analysing interactions with drivers, locations, and even time! Here’s…
Lyft used graph embeddings to improve riding experience for 20M+ riders 🚙 — by analysing interactions with drivers, locations, and even time! Here’s…
Liked by Devesh Raj
Other similar profiles
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 More