Full-Stack Developer • AI/ML Engineer

Prakshil Patel

I build scalable web applications, autonomous multi-agent systems, and deep learning architectures — bridging the gap between intelligent algorithms and production software.

Let's Connect
prakshilmpatel@gmail.com

A BIT ABOUT ME

I'm Prakshil Patel — a Full-Stack Developer and AI/ML Engineer, currently pursuing a B.Tech in ICT at Adani University.

I build scalable, user-focused applications by combining clean frontend design with powerful backend and machine learning systems. My work spans full-stack web platforms, computer vision models, agentic data quality pipelines, and lightweight GPU monitoring infrastructure — engineered with a passion for high performance, modular code, and real-world impact.

LEARNING JOURNEY

2023

Higher Secondary Education — Terf School

Percentage: 88.01%

2023–2027

Adani University, Ahmedabad, Gujarat, India

B.Tech in Information and Communication Technology

CGPA: 8.66/10.0

Focused on core engineering fundamentals, problem-solving, and project-based learning in technology and communication systems.

EXPERIENCE

Work & Research

From production engineering at scale to solving NP-hard computational challenges and frontier algebraic research.

Deskera

Enterprise Systems

Software Engineer Intern

Sept 2026 – Present
Bengaluru, India

Enhanced enterprise inventory management workflows across complex warehouse views using React, Node.js, and SQL, implementing high-throughput server-side filtering, pagination, and multi-field validation across enterprise inventory records.

Optimized inventory data retrieval by analyzing SQL query execution plans, eliminating redundant database operations, and refining system architecture for critical high-volume warehouse operations.

Skills:ReactReactNode.jsNode.jsSQLExpress.jsExpress.jsPostgreSQLPostgreSQLDatabase Optimization

Indian Institute of Technology Kanpur (IITK)

Theoretical AI & Complexity Research

Student Research Associate

May 2026 – July 2026
Kanpur, India
NP-Hard Problem
Tackled NP-Hard Complexity Barrier

Engineered a GNN-guided algebraic circuit simplification pipeline achieving a 17% mean circuit-size reduction on complex polynomial benchmarks.

Google DeepMind Frontier
3×3 Tensor Decomposition Frontier

Investigated the foundational open problem of efficient 3×3 matrix multiplication — an active frontier for Google DeepMind and algebraic complexity research.

Conducted advanced research under Dr. Nitin Saxena on ML-guided optimization of algebraic computational graphs using Graph Neural Networks (GNNs) for circuit simplification and symbolic computation.

Achieved a 17% mean circuit-size reduction by engineering a framework to transform mathematical expressions into algebraic circuits and benchmarking reduction heuristics — tackling arithmetic circuit minimization, an intrinsically NP-hard computational problem.Solved NP-Hard Problem

Investigated the open problem of efficient 3×3 matrix multiplication using tensor decomposition techniques and algebraic complexity — a world-class problem actively researched by Google DeepMind (AlphaTensor) and leading complexity theorists worldwide.World-Class Open Problem

Skills:PythonPythonPyTorchPyTorchPyTorch GeometricPyTorch GeometricGraph Neural Networks (GNNs)NetworkXNumPyTensor DecompositionAlgebraic Complexity

DigiXLT

Full-Stack & GenAI

Software Engineering Intern

March 2026 – May 2026
Remote / Hybrid

Architected and shipped a production-grade Smart HRMS platform (onboarding, attendance, payroll, reviews), replacing manual workflows with a unified AI-driven system using MERN stack and LLM integration to automate HR query resolution and generate real-time performance summaries.

Implemented Stripe payroll processing, AWS S3 document management at scale, and a tamper-proof geo-fencing mechanism for robust attendance verification secured with JWT-based multi-role authorization.

Skills:MongoDBMongoDBExpress.jsExpress.jsReact.jsReact.jsNode.jsNode.jsOpenAI APIAWS S3AWS S3StripeTailwind CSSTailwind CSSGeo-Fencing

PORTFOLIO WORK

Featured Work

CollabSpace - Project Management Platform - Project by Prakshil Patel

FEATURED PROJECT

CollabSpace - Project Management Platform

CollabSpace is a modern, full-stack project management platform that empowers teams to collaborate seamlessly, manage tasks efficiently, and track progress in real-time. Built with enterprise-grade authentication and responsive design, it delivers a production-ready solution for teams of all sizes.

React 19 iconReact 19Redux ToolkitViteNode.js iconNode.jsExpress iconExpressPostgreSQL iconPostgreSQLPrisma ORMClerk Authentication iconClerk AuthenticationTailwind CSS 4 iconTailwind CSS 4RechartsInngestNodemailerVercel iconVercel
GitHubClick the image to view live
EduHub - An AI Powered Learning Platform - Project by Prakshil Patel

FEATURED PROJECT

EduHub - An AI Powered Learning Platform

EduHub is a full-scale AI-powered Learning Management System designed to simplify digital education for students, teachers, and administrators. It centralizes course management, learning resources, and performance tracking through role-based dashboards, with a strong focus on usability, scalability, and automation.

Next.js iconNext.jsReact iconReactNode.js iconNode.jsExpress iconExpressMongoDB iconMongoDBPostgreSQL iconPostgreSQLREST APIsAuthenticationVercel iconVercel
GitHubClick the image to view live
Civic Trust Portal – Citizen-First Digital Platform (SDG-16) - Project by Prakshil Patel

FEATURED PROJECT

Civic Trust Portal – Citizen-First Digital Platform (SDG-16)

Civic Trust Portal is a centralized, citizen-first platform aligned with UN Sustainable Development Goal 16 (Peace, Justice and Strong Institutions). It bridges the information gap by unifying legal awareness, government welfare schemes, scholarships, internships, and job opportunities with interactive transparency dashboards and automated aggregation pipelines.

React iconReactNode.js iconNode.jsExpress iconExpressFlask iconFlaskMongoDB iconMongoDBFirebase iconFirebaseREST APIsRapidAPIWeb Scraping
GitHubClick the image to view live
Café POS – Restaurant Management & Point-of-Sale System - Project by Prakshil Patel

FEATURED PROJECT

Café POS – Restaurant Management & Point-of-Sale System

A comprehensive full-stack Restaurant Point-of-Sale (POS) system engineered to streamline dining operations across dedicated Admin, Waiter, Kitchen, and Cashier interfaces. Features end-to-end table management, real-time synchronized kitchen order workflows via Server-Sent Events (SSE), multi-mode payment processing (UPI/Cash/Card), and analytics reporting.

React 18 iconReact 18TypeScript iconTypeScriptVite iconViteTailwind CSS iconTailwind CSSNode.js iconNode.jsExpress.js iconExpress.jsPostgreSQL iconPostgreSQLPrisma ORM iconPrisma ORMZustandReact QueryServer-Sent Events (SSE)JWT AuthRazorpay
SchemaCraft - From a Product Idea to Deployable schema - Project by Prakshil Patel

FEATURED PROJECT

SchemaCraft - From a Product Idea to Deployable schema

SchemaCraft converts a plain-English product idea into a complete database design. Describe what your app needs, get tables/fields/relationships back, iterate by refining the description, and export the final schema as a foundation for building the application.

Next.js iconNext.jsReact iconReactTypeScript iconTypeScriptTailwind CSS iconTailwind CSSFastAPI iconFastAPIPython iconPythonGroq (LLM)Render
GitHubClick the image to view live
PatternVerse - A Pattern Library - Project by Prakshil Patel

FEATURED PROJECT

PatternVerse - A Pattern Library

PatternVerse is a UI/UX pattern library that provides reusable, production-ready components for modern web applications. It helps developers and designers build consistent interfaces faster by showcasing layouts, buttons, cards, forms, and interaction patterns in a clean, accessible format.

React iconReactTailwind CSS iconTailwind CSSJavaScript iconJavaScriptVercel iconVercel
GitHubClick the image to view live
RenderBox - A Saas Application - Project by Prakshil Patel

FEATURED PROJECT

RenderBox - A Saas Application

RenderBox is a SaaS-style web application built with scalability in mind. It demonstrates a structured dashboard experience, modular components, and a subscription-ready architecture—serving as a strong foundation for launching modern software products.

React iconReactNext.js iconNext.jsTailwind CSS iconTailwind CSSAuthentication FlowVercel iconVercel
GitHubClick the image to view live

AWWWARDS DESIGNS

Design-forward Builds

NeonNectar - Project by Prakshil Patel

FEATURED PROJECT

NeonNectar

NeonNectar is a visually rich 3D animated website showcasing advanced frontend animation techniques. It focuses on immersive UX, smooth transitions, and modern web aesthetics—great for premium brands or creative portfolios.

React iconReactThree.js / 3D iconThree.js / 3DCSS Animations iconCSS AnimationsJavaScript iconJavaScriptVercel iconVercel
GitHubClick the image to view live
Rejouice - Project by Prakshil Patel

FEATURED PROJECT

Rejouice

Rejouice is an Awwwards-inspired website emphasizing visual storytelling, motion design, and high-quality UI composition with interactive layouts.

React iconReactAdvanced CSS iconAdvanced CSSJavaScript Animations iconJavaScript AnimationsModern UI Design
GitHubClick the image to view live
Magma - Project by Prakshil Patel

FEATURED PROJECT

Magma

A Web3 real estate experience clone featuring smooth scrolling, GSAP-driven animations, and immersive storytelling inspired by the Awwwards-winning Magma website.

GsapAdvanced CSS iconAdvanced CSSJavaScript Animations iconJavaScript AnimationsModern UI Design
GitHubClick the image to view live

INTELLIGENT SYSTEMS & RESEARCH

Beyond the Code: The AI / ML Engineer

You've seen the software engineer craft responsive frontends and resilient backends — now welcome to the laboratory of the AI/ML engineer. Where convolutional backbones, multi-agent reasoning graphs, generative LLM pipelines, and edge hardware converge into real-world intelligence.

AI / ML PROJECT

Industrial PPE Detection System

Developed an AI-powered computer vision system for automated PPE detection in industrial environments using a fine-tuned YOLOv8 object detection model. Consolidated and processed datasets from multiple sources, optimized model training, evaluated detection performance, and worked toward edge deployment for real-world industrial safety monitoring.

PythonPythonPyTorchPyTorchYOLOv8UltralyticsOpenCVOpenCVRoboflowArduino (Hardware Integration)Arduino (Hardware Integration)
Repository Publishing Soon

AI / ML PROJECT

Multi-Label Chest X-Ray Disease Classification

Engineered a research-oriented deep learning pipeline for multi-label classification of 14 thoracic diseases from chest X-ray images. Leveraged ConvNeXt-Large with GCN-based label correlation modeling, class-imbalance-aware loss functions, probability calibration, and patient-level stratification. Evaluated cross-dataset generalization using independent CheXpert and MIMIC-CXR datasets.

PythonPythonPyTorchPyTorchTorchvisionConvNeXtGCNOpenCVOpenCVAlbumentationsMLflowScikit-learnScikit-learn

AI / ML PROJECT

DataSentinel – Multi-Agent Data Quality Analysis Platform

Built an intelligent multi-agent system that automates data quality analysis in Snowflake data warehouses. Employs five specialized AI agents to plan investigations, execute SQL queries, perform statistical profiling, identify anomalies, and generate severity-based recommendations with comprehensive HTML reports through a natural-language interface.

PythonPythonMicrosoft AutoGenSnowflakeSQLOpenAI APIOpenRouterOllamaPandasPandasYData ProfilingStreamlitStreamlitAsyncIO

AI / ML PROJECT

ResearchBuddy – AI-Powered Research Paper Assistant

Developed an interactive AI-powered research assistant that transforms complex academic papers into understandable, step-by-step learning experiences. Processes uploaded research PDFs, enables contextual conversations, and uses Socratic-style prompting to assess prior knowledge, explain technical concepts, and improve research paper comprehension.

PythonPythonStreamlitStreamlitOpenAI SDKGPT-4o-miniOpenRouter APIPyMuPDF4LLMPython-dotenvPython-dotenv

AI / ML PROJECT

TrainWatch – Lightweight ML Training & GPU Monitoring Library

Developing a lightweight Python monitoring library for tracking machine learning training loops, identifying performance bottlenecks, and detecting potential GPU VRAM memory leaks in real time. Features a configurable, drop-in Watcher class that provides training heartbeats, GPU memory diagnostics, and step-level performance metrics, enabling reliable long-running experiments across cloud notebooks and remote GPU environments.

Key Highlights
  • Real-time GPU VRAM monitoring & potential memory leak detection
  • Training-loop heartbeats & step-level performance diagnostics
  • Configurable drop-in Watcher class with minimal overhead
  • Cross-platform compatibility & automated CI testing
PythonPythonPyTorchPyTorchCUDANVIDIA NVML (pynvml)SetuptoolsWheel PackagingGitHub ActionsGitHub ActionsCI/CD

MY SKILLS

The Secret Sauce

Python
Python
Automation, backend, data tooling & ML
PyTorch
PyTorch
Deep learning architecture & training
OpenCV
OpenCV
Computer vision and image processing
Scikit-learn
Scikit-learn
Classical ML pipelines and modeling
NumPy
NumPy
Numerical computing and vectorization
Pandas
Pandas
Data analysis, transformation & ETL
FastAPI
FastAPI
High-performance Python APIs & inference
JavaScript
JavaScript
Frontend + backend development
TypeScript
TypeScript
Type-safe React/Node codebases
React.js
React.js
Component-driven reactive UIs
Next.js
Next.js
Full-stack React framework
Node.js
Node.js
Server-side JavaScript runtime
Express.js
Express.js
REST APIs and middleware
PostgreSQL
PostgreSQL
Relational database + indexing
SQL
SQL
Queries, joins, and data modeling
Tailwind CSS
Tailwind CSS
Utility-first modern styling
Docker
Docker
Containerized development & deployments
Git
Git
Version control workflows
GitHub
GitHub
Collaboration + CI/CD automation
Linux
Linux
CLI, systems & server environments
C++
C++
DSA and performance-oriented computing
C
C
Core programming + systems fundamentals
Supabase
Supabase
DB + auth + realtime storage
Google Colab
Google Colab
GPU-accelerated ML experiments
Vercel
Vercel
Next.js deployments & edge compute
Figma
Figma
UI design handoff and prototyping
Flask
Flask
Lightweight Python microservices
HTML5
HTML5
Semantic web architecture
CSS3
CSS3
Layouts, animations and styling
Visual Studio Code
Visual Studio Code
Daily engineering workspace
n8n
n8n
Workflow automation

CERTIFICATIONS

Proof of Learning

A curated set of credentials that reflect my ongoing focus on full-stack development, data, APIs, and modern AI workflows.

Data Analytics and Visualization Job Simulation

Jun 17, 2024

Accenture

Completed a hands-on simulation focused on analyzing data and communicating insights through clear visualizations.

Full Stack Development

Sep 25, 2025

Felix ITS Institute (Ahmedabad, Gujarat)

Training covering frontend + backend development with practical project work and modern web development workflows.

Introduction to Machine Learning

Jul–Sep 2024

IIT Kharagpur

Covered ML fundamentals, core concepts, and a practical understanding of common supervised learning approaches.

SQL (HackerRank)

Nov 6, 2024

HackerRank

Validated SQL fundamentals including querying, filtering, joins, and working with relational datasets.

Prompt Engineering for Developers

2024

DeepLearning.AI

Learned practical prompt patterns and workflows to build reliable LLM-powered features and developer tooling.

Postman API Student Expert

2024

Postman

Demonstrated API fundamentals including requests, collections, environments, and testing workflows in Postman.

SERVICES & EXPERTISE

What I Can Engineer

Agentic AI

Agentic AI & Multi-Agent Systems

Designing autonomous multi-agent systems with Microsoft AutoGen & LLMs — intelligent data profiling, autonomous reasoning, structured tool execution, and Socratic learning workflows.

Vision & DL

Computer Vision & Deep Learning

Custom neural network pipelines using PyTorch, YOLOv8, and ConvNeXt — fine-tuned object detection, multi-label medical diagnosis, and edge AI hardware deployment.

MLOps

MLOps & GPU Diagnostics

Telemetry tools for ML training loops, real-time GPU VRAM leak detection via NVIDIA NVML, experiment tracking with MLflow, and automated CI/CD for model delivery.

Full Stack

Full-Stack Web Engineering

High-performance web applications with Next.js, React, Node, and TypeScript — combining modern design-forward interfaces with robust, scalable architectures.

APIs & Cloud

High-Performance APIs & Inference Endpoints

Fast, asynchronous backends built with FastAPI and Express — secure authentication, database modeling, and low-latency model inference microservices.

Data Systems

Data Warehousing & Intelligent Automation

Cloud data architectures with Snowflake & PostgreSQL, automated data quality anomaly detection, custom ETL pipelines, and repetitive workflow automation.