Skip to content
Software Developer at Anuvadini AISambalpur, Odisha · --:-- IST

Building thebackend behindlanguage AI.

I'm Dinesh — software developer at Anuvadini AI, building scalable Node.js services and APIs that bring ML-based language processing and translation into production web platforms.

View projectsExperienceDownload résuméContact
fig.01translation requestschematic
POST /translate
{ "text": "hello", "to": "or" }
200→ନମସ୍କାରOdia
(01)About

About Dinesh

I work on the layer most people never see — the services, APIs and data flows that keep a product fast, secure and reliable. At Anuvadini AI, that layer powers AI-driven language processing and translation. At AICTE, it served education platforms used by institutions across India.

~/dinesh — zsh
~ $ whoami --verbose
name
Dinesh Kumar Sahoo
role
Software Developer
org
Anuvadini AI
base
Sambalpur, Odisha, IN
local
--:-- IST (UTC+05:30)
degree
B.E. CSE · Chandigarh University
offline
biking · travel · cooking
~ $

What I focus on

  1. 01

    Backend services & APIs

    Scalable Node.js and Express services, REST APIs, and the middleware that keeps them secure.

  2. 02

    ML in production

    Integrating ML-based modules into real products, with hands-on work in computer vision and classical ML.

  3. 03

    Full-stack delivery

    React interfaces through to MySQL data layers, shipped in Agile sprints with peer code review.

Off the clock: open-source projects & new technologies, adventure biking & traveling, cooking and experimenting.

(02)Experience

From enrollment systems to language AI.

Three roles, one throughline: dependable systems people rely on — a university enrollment platform, education tools used by institutions across India, and production language AI.
  1. Jan 2026 — Present Now

    Anuvadini AI

    Software Developer

    India

    Building AI-powered applications for language processing and translation.

    • 01.1Develop scalable backend services and APIs in Node.js that integrate ML-based modules.
    • 01.2Improve performance, reliability and user experience across web platforms.
    • 01.3Work with cross-functional teams to deploy and maintain production-level applications.
    Node.jsREST APIsML integrationLanguage processingTranslation systems
  2. Aug 2025 — Jan 2026

    AICTE

    Software Developer Intern

    All India Council for Technical Education · Internship · India

    Web applications behind national-level education initiatives.

    • 02.1Built and maintained web applications supporting national-level education initiatives.
    • 02.2Worked across the full stack, from frontend interfaces to backend APIs.
    • 02.3Strengthened performance, security and scalability through targeted optimisation.
    • 02.4Debugged and improved existing systems used by institutions across India.
    Full-stackFrontendBackend APIsPerformanceSecurity
  3. Internship

    PG-Technology

    Web Developer Intern

    A university enrollment system, from React UI to MySQL.

    • 03.1Developed a University Enrollment System with React.js and MySQL, integrating frontend and backend.
    • 03.2Built responsive UI components, optimised database queries and improved load time.
    • 03.3Shipped in an Agile team: sprints, stand-ups and peer code reviews.
    • 03.4Collaborated on API integration and debugging, and deployed modules to staging servers.
    React.jsMySQLAPI integrationAgileStaging deploys
(03)Projects

Three builds across vision, security and ML.

Accessibility, trust and mental health: three problems, each solved with a different part of the stack. The figures are live schematics of how each system works. Hover them.
>QWERTYUIOPASDFGHJKLZXCVBNMspaceHAND · 21 LANDMARKS01234567891011121314151617181920INPUT · CAMERAOPENCV → MEDIAPIPE → KEYS
P.01Computer vision · Assistive tech

Virtual Keyboard

Gesture-based typing with Python and OpenCV.

The problem
A physical keyboard isn't accessible to everyone. The goal: a virtual typing interface that lets differently-abled users type hands-free.
How it works
  1. Camera
  2. Mediapipe hand tracking
  3. Finger positions
  4. Virtual keys
  5. Keystroke
What I built
  • Built the gesture-recognition pipeline with Python, OpenCV and Mediapipe hand tracking.
  • Mapped finger positions to virtual keys.
  • Optimised response time for real-time performance.
  • Enabled hands-free typing, applying computer vision to assistive tech.
PythonOpenCVMediapipe
clientReact.jsvalidatemiddlewarebcrypthash · comparejwtsign sessionrole gateaccess control/admindashboard/userdashboardAUTH.LOGROLE: ADMIN→POST /api/auth/login{ email, password }✓validate(body)fields ok✓bcrypt.compare(password, hash)match✓jwt.sign({ sub, role })token issued✓authorize("admin")access granted←200 → /admindashboardtoken eyJhbGciOiJIUzI1Ni….eyJyb2xlIjoiYWRtaW4iLCJz….kX8vN2qLr0Tz…
P.02Security · Full-stack

Authentication System

Full-stack, secure authentication with role-based access.

The problem
Nearly every app needs sign-in, and getting it wrong is costly. The goal: an end-to-end flow that keeps passwords safe and access properly scoped.
How it works
  1. React client
  2. Validation middleware
  3. bcrypt
  4. JWT session
  5. Role-based dashboards
What I built
  • Built the end-to-end flow with React.js on the frontend and Node.js/Express.js on the backend.
  • Implemented JWT session handling and bcrypt password hashing.
  • Wrote middleware for input validation and access control.
  • Built role-based dashboards for admins and users.
React.jsNode.jsExpress.jsJWTbcrypt
01 CLEAN02 FEATURES03 REDUCE04 TRAIN05 VALIDATEREDUCED FEATURE SPACECLASSIFIERSSVMRandom ForestNeural NetworkCROSS-VALIDATIONtrainheld-out foldclass 0class 1
P.03Machine learning

Depression Detection System

ML-based mental health prediction.

The problem
Mental health prediction depends on messy, incomplete data. The goal: a clean pipeline and classifiers that are validated properly, not just fitted.
How it works
  1. Raw data
  2. Cleaning
  3. Feature engineering
  4. Dimensionality reduction
  5. Classifiers
  6. Cross-validation
What I built
  • Preprocessed and cleaned datasets with pandas and NumPy, handling missing values.
  • Engineered features and reduced dimensionality to improve model performance.
  • Trained SVM, Random Forest and Neural Network classifiers with scikit-learn and TensorFlow.
  • Used cross-validation to improve accuracy and recall.
PythonpandasNumPyscikit-learnTensorFlow
(04)Stack

Tools I use, and where I've used them.

Everything here comes from my résumé, and most of it links to where I used it: a role, a project or a certification. Pick one to see the evidence.

  1. L01

    Interface

    What people touch

  2. L02

    Services

    Where requests are handled

    Where Node.js shows up

  3. L03

    Intelligence

    Models and perception

  4. L04

    Data

    Where state lives

  5. L05

    Tooling

    How work gets shipped

  6. L06

    Foundations

    What everything rests on

(05)Education

Foundations.

A computer science degree, and the certifications I picked up along the way: cloud, web, databases and the MERN stack.

Education

  1. 2021 — 2025

    Chandigarh University

    CGPA 7.67 / 10

    B.E. in Computer Science Engineering

  2. 2018 — 2020

    College of Basic Science and Humanities, Bhubaneswar

    Class XII · Council of Higher Secondary Education, Odisha

  3. 2013 — 2018

    Rtapalli Vidyapitha, Bhubaneswar

    Class X · Board of Secondary Education, Odisha

Certifications

  • IBM Skills NetworkC.01

    Introduction to Cloud Computing

  • IBM Skills NetworkC.02

    Web Development with HTML, CSS, JavaScript

  • MongoDB UniversityC.03

    Introduction to MongoDB

  • UdemyC.04

    MERN Stack Development (React.js & Node.js)

(06)Contact

Say hello

Have a project in mind, or a question about my work? My inbox is open.

Based in Sambalpur, Odisha. It's --:-- there right now.

dineshsahoo0406@gmail.com