Machine Learning Engineer @ FiVerity
Built and scaled production LLM systems for fraud-risk summarization with explainability, delivering agentic investigation workflows, and engineering automated threat-intelligence pipelines.
Python
PyTorch
FastAPI
React
LangChain
Docker
EKS
Terraform
HuggingFace
Full-Stack Software Engineer (Data Platforms) @ Supply Trace
Led end-to-end development of the Supply Trace platform from concept to production, translating real-world supply-chain problems into full-stack web applications that enabled supplier–customer network analysis.
React
Node.js
Django
Python
Apache Spark
Kafka
Airflow
Docker
AWS
Data Engineer @ TCS R&D
Engineered IaC-driven data pipelines and established production-grade observability pipelines to support marketing analytics, segmentation, and ML model deployment across multiple enterprise clients.
Python
Airflow
Terraform
AWS ECS
S3
MLflow
Docker
CI/CD
Project Tracker is an Application available for iOS, macOS and iPadOS which helps to track all your projects smartly.
- Stored user data in cloud for ubiquity
- Identifying Patterns from User Interaction
- On Device Recommendation System
- Adapting Face Unlock
SwiftUI
WidgetKit
Core ML
CloudKit
Content Based Filtering
Text miner is an Application available for iOS, iPadOS and macOS which helps in bringing Insights out of Text.
- Handling API Responses
- Implementing OAuth
- Creating Charts and Animations
- Increasing App Interaction by donating intents
SwiftUI
Natural Language
Core ML
SiriKit
Summarization
Sentiment Analysis
Multi Modal Sentiment Classification
Dependency between modes (text, audio and video) is taken into account for obtaining sentiment.
- Extract unimodal contextual representations using self attention and Crossed Attention representations using cross attention layers.
- The generated features are fed to GRU resulting in Deep multi modal contextual representations.
- These representations for each utterance are concatenated and fed to Fully Connected layers followed by softmax layer for final sentiment.
Natural Language
Pytorch
Sentiment Analysis
UnderWater Sonar Image Classification using AW-CNN
CNN is widely known for its application in image classification. However, there are some drawbacks even for this algorithm. So the application of CNN in every field has a variant of it.
- AW-CNN is a variant of CNN solving random initialisation problem of CNN
- AW-CNN uses feature matrix constructed from DBN to initialise the CNN
- The accuracy of AW-CNN was 4% higher than CNN without any overfitting
CNN
Pytorch
Tensorflow
Keras
Data Science
Data Preparation, EDA, Modelling, Model Evaluation and Model Deployment
iOS Development
Xcode, CoreData, UIKit, SwiftUI, CoreLocation, RestFul API management, CoreML, CreateML
Cloud Based Technologies
AWS, Docker, Kubernetes
Web Development
HTML, CSS, BootStrap, React JS, JavaScript, PHP, Node JS
Natural Language Processing
Language Modelling, Named Entity Recognition, Entity Linking, Question Answering