About
With over 10+ years of experience in data analytics, I am passionate about leveraging…
Contributions
-
Your machine learning model's predictions are off target. How will you navigate this unexpected outcome?
Refining models may not always mean redevelopment as there might be different possible reasons for off predictions - You can check the character stability index for the features and if there is a drastic change in any of them, try excluding that feature/s and retrain the model. - Population stability is another factor, you might see the screw and you can try redefining the thresholds. - You can have a better Out of time(retro & recent) sample to mitigate the possibility of the model being deteriorated in a short period of time. - If any of the above is not working, better to go back to the drawing board (redevelopment!) and try to increase the feature types and check for alternate algorithms with a wide range for hyperparameter tuning.
-
Struggling with limited resources in a machine learning project?
Below strategies will help optimize resource use without/ limited compromising performance : 1. Focus on key variables to reduce computation. Feel free to enrich your dataset with third-party vendors to have better visibility/spread of variables. 2. Use techniques like synthetic data generation to expand datasets 3. Use Efficient and resource-friendly models like XGBoost or LightGBM. 4. Utilize scalable cloud platforms for flexible resources (just keep a track of billing) 5. Adapt pre-trained models to save time and computational power. 6. Use random search or Bayesian optimization for Efficient hyperparameter tuning 7. Leverage frameworks like Apache Spark for parallel processing.
-
You're facing a stakeholder dilemma: prioritize model performance or data privacy?
To balance model performance with data privacy we should follow below - Use Privacy-Preserving Techniques: Apply differential privacy, federated learning, or homomorphic encryption to protect data during model training. Leverage Synthetic Data: Train models on synthetic or anonymized data to avoid exposing real user information. Minimize Data Usage: Collect only necessary data to reduce privacy risks and improve model performance. Conduct Privacy Audits: Regularly review data practices to ensure compliance with privacy laws like GDPR or CCPA. Align Stakeholder Expectations: Set clear guidelines on performance vs. privacy trade-offs and communicate regularly with stakeholders.
Honors & Awards
-
Master Stroke
Bajaj Finserv
For Increasing Accuracy of Product Recommendation and Channel Affinity Models
-
GEM (Going Extra Mile)
Bajaj Finserv
-
GEM (Going Extra Mile)
Bajaj Finserv
-
HOM(Hero Of The Moment)
Bajaj Finserv
Award for Exceptional work done during AOP.
-
Finalist
ASIA PACIFIC & MEA CUP 2011
Languages
-
English
Professional working proficiency
-
Marathi
Native or bilingual proficiency
-
Hindi
Professional working proficiency
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 MoreOthers named Aditya K. in India
-
Aditya K.
Director at Google
-
Aditya Varma
Strategy and Transformation Leader | Driving Innovation | MBA | Project Management
-
Aditya k
-
Aditya K
SAP PP/QM ASSOCIATE CONSULTANT | S4 HANA | PRIVATE CLOUD | PP | PP-PI | MTO | MTS | RISE WITH SAP | Digital Transformation | MIGRATION | ROLLOUT | QUALITY MANAGEMENT | QM
-
Aditya K
Lead Full Stack Engineer at Mad Street Den
750 others named Aditya K. in India are on LinkedIn
See others named Aditya K.