My work spans machine learning research, full-stack development, and industrial systems, focusing on bridging the gap between advanced engineering and functional product design
- π¬ AI & Machine Learning Research: Developing hybrid architectures like CNN-Transformer-BiLSTM for clinical wearables and PINNs for solving fluid dynamics and Custom trading algo
- π± Mobile & Web Development: Building high-performance platforms using Flutter and React, including AI-powered agricultural tools and food-ordering apps
- π οΈ Systems & UI Design: Architecting real-time monitoring interfaces for Solar EV systems and large-scale system loggers
- π¨ Product Design: Leveraging the Adobe Suite and Figma to create high-fidelity landing pages and industrial dashboards
- π Based in Bhubaneswar, India: Currently pursuing a B.Tech in Computer Science at KIIT University
β‘ I enjoy turning complex technical ideas into elegant, usable products, like to work with Ml and gen ai. ---
Research Paper Accepted
"Stock Earnings Forecasting via News Factor Analyzing Model" Accepted in Elsevier Procedia Computer Science Selected for presentation at ICMLDE 2025
Hackathon
Qualified for Round 2 β Nirman 4.0 Hackathon Silicon University
Community
Active KSRC Member for over 2 years
| Project | Description | Tech | Stars |
|---|---|---|---|
| Triton Fuser | Hydrodynamic Limit Order Book PINN trained on Binance L2 data, utilizing a C++ AVX2 SIMD ingestion engine (14ns/tick) and a custom OpenAI Triton fused GEMM + Softplus kernel for NVIDIA T4 GPUs. | Python, PyTorch, OpenAI Triton, C++, AVX2 SIMD, CUDA |  |
| Dynamic Gradient PINNs | Developed a robust ResSIREN-Fourier PINN utilizing PCGrad, RAR, and Softplus Physics Anchor balancers to stabilize Navier-Stokes simulations, achieving a 32,700x forward inference speedup over OpenFOAM with near-perfect convergence (<1%). | Python, PyTorch, PINNs, Machine Learning, Neural Networks |  |
| AI Clot Monitoring | Developed the SWAT Tier clinical-grade wearable system utilizing a Bayesian triple-encoder hybrid (CNN-Transformer-BiLSTM) and WGAN-GP data synthesis to achieve 86% Critical Recall in predicting post-surgical blood clots. | Python, PyTorch, CNNs, LSTMs, Machine Learning, Deep Learning |  |
| Sys Logger | High-throughput system observability platform using Next.js and Flask. Architected a real-time telemetry pipeline managed via PM2 to monitor, log, and ingest millions of data points from 100+ endpoints for ML anomaly detection. | Python, Flask, Next.js, PostgreSQL, PM2, Machine Learning |  |
| Krishi Sahayak | AI-powered Flutter agricultural platform with CNN-based crop disease detection, TFLite edge inference, weather intelligence, and ML-driven sowing recommendations via FastAPI, Supabase, and Firebase. | Python, FastAPI, Flutter, CNN, TFLite, Supabase, Firebase |  |
| FinReport | Explainable stock earnings forecasting using NLP-based sentiment analysis on financial news and quantitative indicators; research presented at the ICMLDE Conference. | Python, Pandas, Scikit-Learn, NLP, Machine Learning |  |
| CamTrak | Python-based computer vision application allowing users to control a computer mouse cursor and perform clicks using hand gestures via a webcam to assist disabled patients. (In development) | Python, OpenCV, MediaPipe, Computer Vision |  |
| Road Damage Detector | Computer vision system detecting road surface damage for infrastructure monitoring and predictive maintenance. | Python, OpenCV, Machine Learning, Computer Vision |  |
Outside engineering I work on automotive concept art and digital posters.
- Automotive poster compositions
- Concept vehicle visualization
- Diecast model collection
Design for me is not separate from engineering. It is simply engineering with aesthetics.
"Elegance in engineering, clarity in design, precision in systems."