bFan
CompleteA streamlined smart-device control experience focused on simple setup and reliable everyday operation.
Building AI platforms, connected applications, and scientific embedded systems at Pashupatastra Solutions — across multi-model AI, STM32/AD5941 hardware, quality-control tools, and biosensor workflows.
I'm a Software Developer at Pashupatastra Solutions, joined through campus placement after completing my B.Tech in Computer Science & Engineering (AI) at Purnea College of Engineering, Bihar Engineering University (CGPA 8.81). Before that, a Diploma in Electronics Engineering — which is where the IoT and embedded side of my work comes from. I like owning the full stack of a problem: training the model, building the API and UI around it, and — when the problem is physical — wiring the sensor that feeds it. At Pashupatastra Solutions, that range now spans six active product initiatives: bFan, Oncogon AI, Multi-Model, UPR Manager, Strip Tech, and AddiPRECISE. Public portfolio descriptions remain intentionally high-level to protect proprietary architecture, client information, research data, and internal workflows. My earlier work includes two skin-cancer classifiers, a production marketplace platform, and an IoT EV charging monitor that won a state hackathon.
Ensemble deep learning model (DenseNet201 + Inception V3) classifying seven skin disease categories from dermoscopic images. Built as an evolution of an earlier binary classifier into a scalable framework for AI-assisted clinical screening and early diagnosis support.
Real-time IoT monitoring system for EV wireless charging, covering energy tracking, grid integration, and remote control via the Blynk platform.
Full-stack marketplace with role-based access control (Admin/Buyer), Razorpay + manual UPI payment gateway, and digital order fulfillment.
Logistic regression classifier trained on breast cytology data, achieving 96.4% accuracy and validated via ROC-AUC and confusion matrix.
CNN-based system for binary (Benign/Malignant) skin lesion classification, built to validate the feasibility of AI-driven medical imaging before scaling up to a multi-class model.
Production-ready student registration platform built for SVN Infra & Solar Service Pvt. Ltd., supporting online enrollment and payment processing.
Multi-mode home appliance control via Google Assistant, Alexa, IR remote, and manual switch — no dedicated smart hub required.
Pashupatastra Solutions · selected initiatives
A streamlined smart-device control experience focused on simple setup and reliable everyday operation.
A research-focused AI platform supporting structured scientific data and responsible analytical workflows.
A controlled multi-model AI workspace designed to support research, engineering, and automation tasks.
An offline-capable engineering and quality-control workspace for connected measurement hardware.
Research and engineering work for a reusable electrochemical sensing platform and its quality workflows.
An operator-facing application ecosystem for configurable electrochemical testing and result workflows.
Confidentiality note: descriptions are intentionally limited to high-level responsibilities and product domains. Proprietary architecture, client information, datasets, implementation details, and internal workflows are not disclosed.