Hi, I'm SOUMYARANJAN SAHOO
Bridging the gap between Hardware, AI, and FinTech. I build secure, scalable systems and high-precision predictive models.
About Me
Engineering the Future of Data
I am an Engineering Undergraduate (B.Tech in ETC) at VSSUT, Burla (Class of 2027), with a deep passion for FinTech and MedTech.
Currently, I am leveraging Machine Learning approaches in the medical field, focusing on high-precision data processing and anomaly detection (e.g., Melanoma and Pneumonia detection). I am also actively transferring this rigorous analytical skillset to the financial sector to handle large time-series datasets and optimize secure, scalable digital payment solutions.
Beyond code, I am an Upcoming Author of two Book Chapters on FPGA hardware acceleration and Agentic AI respectively, and a consistent Kaggle Expert in Medical Image Analysis.
Technical Toolkit
Languages
Python, Java, SQL, Verilog, Dart (Flutter)
AI & ML
TensorFlow, Scikit-Learn, OpenCV, EfficientNet, Transfer Learning, Hugging Face
Hardware
VLSI Design, FPGA (PolarFire SoC, Xilinx), Embedded Systems
Web & Cloud
Flask (Backend), Nginx, Hugging Face, IBM Cloud, REST APIs
Experience
Research Intern
Engineered an Extra Tree and Random Forest regression pipeline mapping geometric parameters of Cold Form Steel to structural yield stress, mapping outcomes with <5% error rate.
App Development Intern
Contributed to Udra Finance App, integrating a Flutter frontend with a high-performance Python Flask backend. Implemented JWT authentication and improved API response times by 40%.
Data Analytics Intern
Conducted comprehensive data analysis to evaluate healthcare accessibility trends using Python. Uncovered correlations between lower income levels and reduced insurance coverage.
AI & Cloud Intern
Architected "Heleophile," a conversational AI chatbot on IBM Cloud to accelerate solar energy adoption. Engineered logic to calculate installation costs and ROI based on real-time data.
Featured Engineering Work
Derma-Vision: Wiener Filtering and Artifact Restoration for High-Fidelity ML Prediction
A core module of the Diagnomass ecosystem. Engineered a Dual-Stream EfficientNet-B7 ensemble fusing raw RGB pixels with offline-computed CTE (Color-Texture-Edge) maps and tabular patient metadata via a custom MLP. Deployed as a full-stack clinical web dashboard.
0.9941 Validation AUC & 0.9659 F1-Score via anatomically stratified testing.
PneumoScanAI
Advanced CNN architecture for Pneumonia Detection in Chest X-Rays. Engineered an automated pipeline with image normalization, Adam optimization, and data augmentation.
State-of-the-art accuracy of 97.86%.
DeepGuard (Deepfake Detection)
Architected the backend infrastructure for a cross-dataset robust ensemble model for video deepfake detection. Deployed inference pipeline on Hugging Face.
Real-time video API analysis (Paper Under Review Feb '26).
Udra Finance App
A secure FinTech logistics maintenance application. Developed Python Flask backend and a Flutter frontend. Managed environments, Nginx server deployment, and JWT authentication.
Streamlined complex payment logistics tracking.
Real-Time Face Tracking (FPGA)
Hardware implementation of a visual tracking pipeline on the Microchip PolarFire SoC Video Kit.
Utilized FPGA fabric to achieve hardware-accelerated real-time frame rates.
Scientific ML for CFS
Developed a Scientific ML framework at IIT Hyderabad to map complex geometric parameters of Cold-Formed Steel to structural yield stress.
Substituted expensive lab testing with accurate predictive equations.
Honors & Achievements
Book Chapter Author(Accepted - Mar 2026)
Chapter 15: "The Sovereign Agent: A Framework for Fusing Agentic AI with Decentralized Systems" — Agentic Artificial Intelligence for Automation and Decision-Making, Scrivener Publishing (Wiley). Explores Distributed Intelligence, Device Agent frameworks, and the ETHOS framework.
Book Chapter Author (Submitted to the Publisher - Feb 2026)
Chapter 5: "Methodologies for Implementing Complex AI Models on FPGA Systems" — FPGA for Next Generation IC Design and Autonomous Systems, Scrivener Publishing (Wiley). Explores the Details of hardware acceleration of CNNsand their implementation on FPGAs.
2nd Prize, Student Poster Contest 2025
Organized by the IEEE Solid-State Circuits Society (SSCS), Kolkata Chapter.
Rank 38 (out of 247 teams)
AI/ML-Powered FPGA Design & Simulation Hackathon (2025).