y = σ(Wx + b)
∇L = ∂L/∂w
J(θ) = -Σ y log(ŷ)
softmax(z) = e^zi / Σe^zj
ReLU(x) = max(0, x)
Open to research internships & ML systems work

Hi, I'mAmisha👋

I work on end-to-end machine learning systems—from messy, real-world data and modelling, through evaluation and iteration, to deployment-ready APIs and dashboards.

Lately, that has meant wildfire early warning, road-safety analytics, and clinical decision support—projects where careful modelling and solid engineering can actually move the needle.

📊
4+
Research projects
🏆
3
Major recognitions
💻
2+
Years in ML
View projectsGet in touch
AI / ML
Real-world
Neural model online
🤖

Abstract representation of an AI engineer orchestrating data, models, and systems.

Training models…

Monitoring metrics

Training progress
🚀
💡
⚡
🎯

About me

Get to know the person behind the projects.

Research-focused developer at work

Hi there! 👋

🚀
💡

A bit about me

I'm an Integrated MSc Mathematics student at NIT Rourkela, with a strong interest in how mathematical ideas, data, and machine learning come together in real systems. I enjoy turning vague, noisy problems into clear pipelines, experiments, and tools that others can rely on.

I'm generally excited about any system work where ML, data, and thoughtful engineering come together – not just one narrow domain. Spatiotemporal forecasting, risk modelling, and domain-specific retrieval systems are things I've especially enjoyed recently.

Outside of projects, you'll usually find me experimenting with new model ideas, improving my problem-solving skills, or reading about how ML systems are deployed and maintained in the real world.

End-to-end systems

Research-driven

Always learning

Research interests

What I'm excited about exploring and building next.

  • Core machine learning and deep learning: representation learning, optimization, generalization
  • Applied ML across domains – vision, time-series, tabular, and multimodal data
  • Generative and probabilistic models that can reason under uncertainty
  • ML systems: data pipelines, evaluation, monitoring, and deployment in the real world
  • Responsible and interpretable AI: robustness, failure modes, and human-centered evaluation
  • Search, retrieval, and recommendation systems powered by modern embeddings and RAG

ML & Statistical Modelling

  • Supervised & unsupervised learning
  • Deep learning (PyTorch)
  • Time-series & sequence models
  • Evaluation & error analysis

Data & Systems

  • Python, C/C++
  • Data structures & algorithms
  • SQL & analytical queries
  • ETL over geospatial & sensor data

Serving & Tooling

  • FastAPI & REST services
  • Docker & deployment basics
  • Experiment tracking
  • Dashboards & storytelling

Stack I use

Technologies I work with to build machine learning models and systems that solve real problems.

Python

JavaScript

C++

Java

R

Structured Query Language

TensorFlow

PyTorch

Keras

Scikit-learn

React

Next.js

FastAPI

Flask

Pandas

NumPy

Plotly

OpenCV

MongoDB

PostgreSQL

Amazon Web Services

Google Cloud Platform

Microsoft Azure

Docker

Kubernetes

Git

GitHub

Visual Studio Code

Jupyter

Linux

Python

JavaScript

C++

Java

R

Structured Query Language

TensorFlow

PyTorch

Keras

Scikit-learn

React

Next.js

FastAPI

Flask

Pandas

NumPy

Plotly

OpenCV

MongoDB

PostgreSQL

Amazon Web Services

Google Cloud Platform

Microsoft Azure

Docker

Kubernetes

Git

GitHub

Visual Studio Code

Jupyter

Linux

Research projects

View more on GitHub
PythonU-NetGoogle Earth Engine

Spatiotemporal Wildfire Early Warning (Agnirhodhak)

Lead ML developer

Forecast wildfire risk over satellite and sensor data to surface high-risk regions days in advance.

PythonU-NetGoogle Earth EngineGeospatial ML
Code
Time-seriesRisk modellingDashboards

SafeRoadAI – Road-safety Analytics

ML engineer

Rank high-risk road segments from traffic and crash data to guide targeted, data-driven interventions.

Time-seriesRisk modellingDashboards
Code
Audio MLLibrosaCNNs

AuscultoML – Lung Sound Disease Classification

Research project lead

End-to-end audio ML pipeline that classifies lung sound recordings into disease categories for decision support.

Audio MLLibrosaCNNsFastAPI
Code
RecommendersRankingCompetition

Amazon ML Challenge – Smart Product Pricing

Top 300 / 82,787 participants

Built and tuned ranking models for price recommendations as part of the Amazon ML Challenge 2024.

RecommendersRankingCompetition
Code

Spatiotemporal wildfire early warning

A live demo of my Agnirhodhak project: forecasting wildfire risk over satellite and sensor data to surface high-risk regions in advance.

Open live demo

More from GitHub

See all repositories

Loading repositories...

Publications & achievements

  • Amazon ML Challenge 2025

    Top 300 out of 82,787 participants for price recommendation models.

  • BookingJini March Cohort Hackathon Winner (2025)

    Won cohort hackathon for building data-driven product features.

  • Google Developer Groups (GDG) On Campus Recognition (2025)

    Recognized for contributions to ML workshops and student developer community.

Experience & education

WSN Lab

Wireless Sensor Networks Lab, NIT Rourkela

Summer Research Intern (ML)

May 2025 – Jul 2025

Technical contributions

Built U-Net based pipelines, geospatial ETL, and inference services over large-scale satellite data.

Impact

Focused on building robust spatiotemporal pipelines over satellite and sensor data to support data-driven decision making.

U-NetGeospatial processingTime-seriesPython

2022 – 2027 (Expected)

Integrated MSc in Mathematics

National Institute of Technology Rourkela

Rourkela, Odisha

Let's connect

I'm actively looking for ML research internship opportunities.

Prefer direct links? You can also reach me via:

Email GitHub LinkedIn