~/projects

Projects

An engineering ledger mapping data pipelines, algorithmic layers, and metric validations.

German Aerospace Center (DLR)2025

Semantic Segmentation of CFRP Microstructures

Computer VisionDeep LearningU-NetImage ProcessingPython

Ingested Data Pipeline

High-resolution material micrographs of Carbon Fiber Reinforced Polymers (CFRP).

Algorithmic Engineering

Developed a high-precision semantic segmentation system utilizing specialized U-Net neural network architectures to automate macro-fiber and micro-crack analysis layers.

Systemic Validation

Accelerated material science defect evaluation workflows by replacing manual microstructure annotation bottlenecks with deep learning inference.

BRICS, TU Braunschweig2024

Predictive Modeling for Neurodegenerative Progression

Machine LearningSignal ProcessingScikit-LearnXGBoostRandom Forest

Ingested Data Pipeline

Multi-dimensional, non-invasive biomedical vocal frequency metrics and acoustic signals.

Algorithmic Engineering

Engineered an advanced diagnostic data pipeline using ensemble machine learning models (XGBoost, Random Forest regressors) optimized for tracking neurodegenerative severity.

Systemic Validation

Successfully isolated high-variance vocal diagnostic acoustic signatures to map out Parkinson's disease progression trends with minimal tracking error.

AVIS E-Solutions (Contaque) | Mumbai Metro2023 - 2025

Multi-Lingual RAG Chatbot Architecture

Generative AIRetrieval-Augmented Generation (RAG)LLM Fine-TuningNLPJava

Ingested Data Pipeline

Large-scale, unstructured, multi-lingual municipal enterprise documentation and transit guidelines.

Algorithmic Engineering

Architected production-ready Retrieval-Augmented Generation (RAG) loops and fine-tuned open-weight Large Language Models (LLMs) to output context-aware, deterministic responses.

Systemic Validation

Aggregated real-time systemic API logging architectures to execute deep-dive verification of model performance thresholds and automated system anomaly detection.

AVIS E-Solutions (Contaque) | Indian Railways2023 - 2025

AI/ML Decision-Support Infrastructure

Machine LearningDecision SupportScalable BackendJavaSpring Boot

Ingested Data Pipeline

Real-time telemetry feeds, operational track environments, and locomotive sensory data streams.

Algorithmic Engineering

Engineered high-reliability backend algorithmic layers supporting intelligent decision-support mechanisms to elevate runtime safety margins for locomotive personnel within the "Kavach" framework.

Systemic Validation

Implemented robust Java and Spring Boot backend microservices explicitly configured to handle low-latency decision evaluation natively in highly concurrent environments.

Intellectual Property Portfolio2023

MEDIN - Real-Time Body Vitals Tracking Platform

Computer VisionDeep LearningTime Series AnalysisEmbedded AI

Ingested Data Pipeline

Continuous biometric sensory streams capturing SpO2 saturation profiles, BPM metrics, and locomotion data.

Algorithmic Engineering

Designed an advanced multi-modal sensor fusion metrology pipeline configured to parse diagnostic health signatures from time-series datasets.

Systemic Validation

Protected via dual Indian Government Patent applications: "Miscarriage Prediction System and Method" and "A Complete Metrology of Body Vitals".

Independent Research2023

Intelligent Multi-Modal Document Retrieval System

Natural Language Processing (NLP)Artificial IntelligenceMulti-Modal Search

Ingested Data Pipeline

Unstructured text corpora, nested book matrices, and multi-modal figures embedded within complex PDF documentation.

Algorithmic Engineering

Developed an intelligent end-to-end document parsing and vector embedding ecosystem enabling semantic retrieval matching across unified image and text vector spaces.

Systemic Validation

Enabled context-aware conversational question-answering systems executing across deep text contexts with structural hallucination control.