Computer Vision & Machine Learning

SHAUNAK
MISHRA

ML Researcher

Computer Vision Engineer and ML Researcher. Specialized in medical imaging segmentation, deep learning model benchmarking, and building production fullstack systems.

8.72

SOA CGPA

250+

DSA SOLUTIONS

19

MS AI BADGES

shaunak@portfolio
shaunak@portfolio:~$ welcome
Shaunak Mishra | B.Tech CSE | SOA University
Focused on Deep Learning, Computer Vision & Fullstack Development.
DSA: 250+ problems solved across LeetCode & GFG.
MS Build AI Hackathon 2026: WorkflowOS project.
Type 'help' to see available CLI commands.
shaunak@portfolio:~$
>
SM

Shaunak Mishra

Computer Science & Engineering Scholar • SOA University

ABOUT

ABOUT ME

SM

Shaunak Mishra

SOA University (ITER), Bhubaneswar (CGPA: 8.72)

B.Tech CSE student at SOA University focused on Deep Learning, Computer Vision, and full-stack development.

I'm a Computer Science student passionate about the intersection of machine learning and software engineering. My work spans building deep learning models for medical imaging, developing production-grade web applications, and contributing to open-source projects.

My research focuses on brain tumor classification and segmentation using convolutional and transformer architectures on MRI data. I've benchmarked 14 classification models and 10 segmentation architectures on the BRISC2025 dataset, with a combined MambaOut + SwinSeg pipeline achieving state-of-the-art results.

Tech Stack Focus

PythonC++TypeScriptPyTorchTensorFlowOpenCVUNetReactNext.jsTailwind CSSNode.jsExpressGitDocker

Key Statistics

250+

DSA Solutions

LeetCode & GFG algorithmic keys

14+

ML Architectures

Implemented CV & deep networks

11+

Software Projects

Fullstack platforms & products

19

MS Learn Badges

Academic cloud credentials

Education & Research Timeline

B.Tech Student (CSE)

2023 – 2027

SOA University (ITER), Bhubaneswar

CGPA: 8.72 / 10.0

In-depth specialization in Computer Vision, Advanced Data Structures, Operating Systems, and Relational Databases.

Frontend Developer

Jun 2025 – Aug 2025

CN Nexus — Coding Ninjas Club, SOA University

Improved problem discovery speed by 30% by building topic-wise categorization and memoized search filters for a DSA platform aggregating LeetCode, GFG, and Codeforces

Founding Engineer

Dec 2025 – Mar 2026

Mentomania

Designed and built complete frontend UI/UX for a government exam mentorship platform, delivering full product from design to deployment

RESEARCH

COMPUTER VISION IN MEDICAL IMAGING

14 classification and 10 segmentation architectures benchmarked on BRISC2025 (6,000 T1-weighted MRI slices, 4 classes). MambaOut + SwinSeg pipeline with GradCAM++ explainability.

MambaOut + SwinSeg Pipeline

01

Raw MRI Intake

Ingest high-res DICOM scans spanning T1-CE, T2, and FLAIR profiles.

02

Preprocessing

Apply pixel normalization, non-local means filtering, and spatial resizing.

03

MambaOut Classifier

State-space model classifies tumor type from preprocessed MRI slices.

04

SwinSeg Segmentation

Swin Transformer-based decoder extracts precise tumor boundaries.

05

GradCAM++ Map

Generate activation heatmaps revealing backprop weights on diagnostic regions.

Interactive Grad-CAM MRI Viewer
Simulation
Select Patient MRI

Select a Scan to Begin

Choose a patient scan from the list and click "Analyze Scan" to run the diagnostic pipeline.

Architecture Benchmark (BRISC2025)
MambaOut (Classification)99.50% Acc
Params: 42.3MTest Accuracy
maxViT (Classification)99.50% Acc
Params: 39.6MTest Accuracy
ViT (Classification)99.40% Acc
Params: 45.7MTest Accuracy
ResNeXt-50 (Classification)99.30% Acc
Params: 25.0MTest Accuracy
SwinSeg (Segmentation)0.8765 Dice
Params: 62.4MDice Score
DeepLabV3+ (Segmentation)0.8568 Dice
Params: 43.5MDice Score
UNet++ (Segmentation)0.8178 Dice
Params: 36.7MDice Score
UNet (Segmentation)0.8080 Dice
Params: 31.0MDice Score

PAPER IN PREPARATION

Comprehensive Benchmarking of Deep Learning Architectures for Brain Tumor Classification and Segmentation on BRISC2025

Systematic evaluation of 14 classification models and 10 segmentation architectures on 6,000 T1-weighted MRI slices. The proposed MambaOut + SwinSeg pipeline with GradCAM++ explainability achieves state-of-the-art results under faculty supervision at SOA University.

03 / CREATIVE IMPLEMENTATIONS

ADVANCED ENGINEERING WORK

11

Active Projects

1

OS Contributions

5

Featured Projects

FULLSTACK
Featured

WorkflowOS

FastAPI

AI agent swarm that converts meeting transcripts into execution plans. Microsoft Build AI Hackathon 2026.

FastAPIAzure OpenAINext.jsCosmos DB+3
Explore Architecture
ML-AI
Featured

Brain Tumor Detection

Python

Medical image classification & segmentation research. 14 classification + 10 segmentation architectures benchmarked on BRISC dataset (6,000 MRI scans).

PythonPyTorchTensorFlowtimm+6
Explore Architecture
FULLSTACK
Featured

MentorHub / MentoMania

Next.js

Mentorship platform connecting students with exam toppers (IIT-JEE, NEET, CAT, UPSC).

Next.jsReactTypeScriptTailwind CSS+3
Explore Architecture
FULLSTACK
Featured

AI Vehicle Health Diagnostics

Next.js

AI/ML-based vehicle health diagnostics, predictive maintenance, and driving simulation.

Next.jsFirebaseGenkitTailwind CSS+1
Explore Architecture
ML-AI

Metal Defect Detector

Python

Surface defect detection on metal surfaces using deep learning.

PythonFlaskDeep LearningOpenCV
Explore Architecture
ML-AI

AI Resume Analyzer

React Router

AI-powered resume parsing and scoring with Docker deployment.

React RouterTypeScriptTailwind CSSDocker
Explore Architecture
FULLSTACK

Civic Issue System

React

Platform for civic issue reporting and tracking.

ReactTypeScriptViteTailwind CSS
Explore Architecture
OPEN SOURCE
Featured

openfoodfacts-nodejs

TypeScript

PR #646 merged into v2.0.0-alpha.8. Added getProductImageFolder() to production SDK.

TypeScriptJestNode.jsSDK Development
Explore Architecture
TOOLS

leetsync-organizer

Python

Python tool that auto-categorizes LeetCode solutions into difficulty folders via GitHub Actions.

PythonGitHub ActionsGit
Explore Architecture
TOOLS

DSA-Solve

C++

250+ DSA problems solved in C++ covering Arrays, DP, Graphs, Trees, Binary Search.

C++
Explore Architecture
FULLSTACK

Nexus-E1

TypeScript

DSA platform aggregating LeetCode, GFG, and Codeforces problems.

TypeScriptReactREST APIs
Explore Architecture
SKILLS

TECHNICAL SKILLS

Categories

Currently Exploring

Multi-Modal Vision Transformers & Mamba Architectures

Investigating attention mechanisms in Vision Transformers for medical image analysis and multi-task learning.

MACHINE LEARNING & CV

PyTorch, TensorFlow, and deep learning architectures for computer vision.

PyTorch & Deep Stack95%

Applied: Benchmarked 14 classification models on brain tumor MRI data.

Segmentation U-Net92%

Applied: Segment tumor regions from brain MRI using UNet/UNet++/SwinSeg.

ResNeXt / ViT90%

Applied: Built classification pipelines with ResNeXt, ViT, and MambaOut.

Explainable AI GradCAM++90%

Applied: Generate GradCAM++ heatmaps for model interpretability.

TensorFlow & Keras85%

Applied: Trained deep learning models for classification tasks.

05 / ALGORITHMIC RIGOR

DATA STRUCTURES & COMPETITIVE CODE

/**
 * @file dijkstra.cpp
 * @brief Verified sub-15ms Dijkstra Shortest Path Solution
 */
#include <vector>
#include <queue>

using namespace std;

class ShortestPathSolver {
public:
    vector<int> matchPath(int N, vector<vector<pair<int, int>>>& adj, int src) {
        vector<int> dist(N, 1e9);
        priority_queue<pair<int, int>, vector<pair<int, int>>, greater<pair<int, int>>> pq;
        
        dist[src] = 0;
        pq.push({0, src});
        
        while(!pq.empty()) {
            int d = pq.top().first;
            int u = pq.top().second;
            pq.pop();
            
            if(d > dist[u]) continue;
            
            for(auto& edge : adj[u]) {
                int v = edge.first;
                int weight = edge.second;
                if(dist[u] + weight < dist[v]) {
                    dist[v] = dist[u] + weight;
                    pq.push({dist[v], v});
                }
            }
        }
        return dist; // Time: O(E log V) matched
    }
};
C++17 Compiler • GCC 11.2
Lines: 32
LeetCode Diagnostic Center
Interactive Widget
1+Solved
Easy Level
95 solved
Medium Level
128 solved
Hard Level
30 solved
Algorithmic Subtopic Distribution
Dynamic Programming: 45 solvedGraph Algorithms: 38 solvedBacktracking / DFS: 32 solvedTree Node Traversals: 54 solvedTwo Pointers / Sliding: 44 solvedGreedy Heuristics: 40 solved

DSA-Solve Hub Repository

Holds production implementations of graph traversal, dynamic programming structures, trees maps. Underpinned by speed analysis test units.

ACHIEVEMENTS

MILESTONES & CERTIFICATES

Timeline

2024

Introduction to Software Engineering

IBM via Coursera

View Certificate
2024 & 2025Leadership

Runner-Up, SOA University Chess Tournament

SOA University

Secured runner-up position in university-level chess tournaments two consecutive years.

View Certificate
Dec 2025Certificate ID: 25231210471

Mastering AI and Data Science — Design and Build Intelligent Solutions

IIT (ISM) Dhanbad & TCS iON

Completed intensive course organized by IIT Dhanbad in collaboration with TCS iON. Sep–Nov 2025. Certificate ID: 25231210471.

View Certificate
2025University Level

Smart India Hackathon 2025

Ministry of Education, India

Problem Statement #95 — Participated in SIH 2025 at university level (SOA University).

View Certificate
2026

Microsoft Build AI Hackathon 2026

Microsoft

Built WorkflowOS — AI agent swarm converting meeting transcripts to execution plans. Track: Agent Swarms.

View Certificate
Microsoft Learn Badges
15 Badges

Verified Microsoft Learn credentials spanning AI, security, compliance, and infrastructure.

Learning PathJun 2026

Design a dream destination with AI

Microsoft Learn Verified
Learning PathJun 2026

Introduction to Microsoft security solutions

Microsoft Learn Verified
Learning PathJun 2025

Introduction to security, compliance, and identity concepts

Microsoft Learn Verified
Learning PathMay 2026

Describe the concepts of cybersecurity

Microsoft Learn Verified
ModuleJun 2026

Search and investigate with Microsoft Purview Audit

Microsoft Learn Verified
ModuleJun 2026

Design a dream destination using Microsoft Copilot

Microsoft Learn Verified
ModuleJun 2026

Describe threat protection with Microsoft Defender XDR

Microsoft Learn Verified
ModuleJun 2026

Describe security management capabilities in Azure

Microsoft Learn Verified
CONTACT

GET IN TOUCH

Location
Bhubaneswar, Odisha, India

Bhubaneswar, Odisha, India

Siksha 'O' Anusandhan University Campus Area, Bhubaneswar.