Abhimanyu Bellam
Data Scientist
LexisNexis
Biography
Abhimanyu Bellam brings an excellent amalgamation of industry experience and academic understanding of DS/AI/ML. He has worked on various types of data including images, videos, texts, posts and supply chain data, leveraging the latest technologies to perform computer vision, NLP and forecasts. Furthermore, he has excellent skills in optimizing and coming up with solutions to solve any problem.
His breadth in this field is his confidence, and his skills ensure that he can go into depths as well. He has completed his Master’s thesis on Explaining the Fairness Impact Due to Quantization, and his paper has been accepted to the NeurIPS Workshop on Fine-Tuning Modern Machine Learning: Principles and Scalability, Dec 2024.
On the deployment side, he can use several cloud services like AWS, Azure in conjunction with MLOps, databricks, and without any of these too.
In all, he is an adept engineer, researcher and deployer.
Master’s Courses: Thesis on explaining the disparate impact due to Post Training Quantization, Resource Depenedent Learning, Automated Learning and Data Analysis, Neural Netowrks and Deep Learning, Software Engineering, Object Oriented Deisgn and Development, Statistics, Independent study on Neural Network Calibration and Pruning.
- Artificial Intelligence
- Resource Constrained Neural Networks
- Fairness
- Metaheuristic Optimization
- Algorithms
Master of Science in Computer Science, 2024
North Carolina State University
Bachelor of Technology in Computer Science, 2020
Mahindra University
Skills
Experience
- Developed a pipeline to handle Letter/Email writing capabilities of Lexis+ AI Law Assistant for Canada Jurisdiction
- Handled user law queries, built proprietary workflows to extract legal topics, relevant documents from LexisNexis OpenSearch indexes
- Performed prompt engineering, finetuning, and training custom language models for retrieved document evaluation
- Create balance between inventory and retails to minimize losses using optimization algorithms and ML
- Enhance existing ML and Statistical models
- Enable CI/CD for internal frameworks and applications
- Thesis under Dr. Jung-Eun Kim
- Research the impacts of Neural Networks Quantization on Fairness
- Explain the internal causes of these impacts
- Identify/propose mitigation solutions
- Solved supply-chain issues using ML and Statistics
- Proposed production-timelines for Mack Trucks (a Volvo subsidiary) to meet retail targets
- Performed Gene significance analysis using statistical tests
- Expanded research to use Recurrent Neural Networks for electrical time-series data
- Developed baby detection, covered face and expression identification algorithms
- Utilizeed AWS services to host ML pipelines
- Forecasted sales and customer acquision
- Performed real-time video analytics on CCTV based video inputs
- Used supervised and unsupervised ML and DL algorithms to provide solutions to traffic and Digital Signage related companies.
- Developed facial recognition based attendance systems
- Devised algorithms to identify wrong way driving and OCR for automobiles
- Developed a novel algorithm for the Warehouse Resource Allocation problem, an NP-Hard problem
- Utilized Multi-Objective meta-heuristic algorithms and Divide & Conquer approaches to enhance solutions
- Experimented on Image and video captioning using Convolutional Neural Networks & Recurrent Neural Networks
- Visualized and trained Neural Network architectures using NVIDIA Deep Learning GPU Training System (DIGITS).
- Simulated motion-plans in an autonomous environment considering obstacle avoidance
- Implemented Car Following and Lane Changing traffic models
Certificates
Projects
