Skip to content
View riju-talk's full-sized avatar
💭
Making big changes soon
💭
Making big changes soon

Block or report riju-talk

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
riju-talk/readme.md
Banner

Rijusmit Biswas

ML Engineer · Applied Scientist · Data Engineer · Full Stack Systems Architect

Building intelligent, scalable, production grade systems across ML, data, and modern web platforms.

Portfolio · LinkedIn · Email · GitHub


; Photography;

Original photography — visual thinking beyond code.


Focus Areas

  • Agentic AI & LLM Systems — RAG, tool using agents, evaluation, safety and reasoning
  • Applied ML Research — model behavior, benchmarks, alignment, failure analysis
  • MLOps & Data Engineering — scalable pipelines, deployment, monitoring
  • Full Stack Systems — distributed APIs, real time platforms, end to end ownership

Projects

Research Systems

Project Description
Sickle++ Satellite crop monitoring with training, evaluation, and Earth Engine download utilities. Ingests multi sensor imagery (Sentinel 2, Sentinel 1, Landsat 8) in cell based and plot based formats. Includes UTAE, UNet3D, and ConvLSTM for crop type classification, phenology prediction, and yield estimation.

ML Systems

Project Description
MineMEETS End to end ML pipeline system for data ingestion, feature workflows, experiment tracking, and production deployment patterns.
TURBO Independent implementation of Target Augmented Shared Fusion based Multimodal Sarcasm Explanation. Detects sarcasm, identifies targets, and fuses text and image inputs via modified BART encoder with cross attention and Vision Transformer features. Evaluates on MORE+ dataset with BLEU, ROUGE, and METEOR.

Full Stack Applications

Project Description
DreamDOT Next gen social media and blogging platform that empowers creators to publish, connect, and earn through dynamic subscriptions, pay per view content, and immersive community interaction.
Flourish Smart plant care web application that bridges the gap between wanting to keep plants alive and actually knowing how. Uses real horticultural data combined with artificial intelligence.

🛠 Tech Stack

Languages: Python · TypeScript · Go · C++ · JavaScript
ML & AI: PyTorch · Transformers · RAG · MLflow · NumPy · Pandas
Backend: FastAPI · Node.js · GraphQL · Microservices
Frontend: Next.js · React · Tailwind CSS
Data & Infra: PostgreSQL · MongoDB · Docker · Cloud Platforms


From research to production — I ship systems that reason, adapt, and scale.

Pinned Loading

  1. DreamDOT DreamDOT Public

    A next-gen social media and blogging platform that empowers creators to publish, connect, and earn through dynamic subscriptions, pay-per-view content, and immersive community interaction. Designed…

    TypeScript 3 1

  2. MineMEETS MineMEETS Public

    MIneMeets is an AI-powered multi-modal agent that listens to meetings, transcribes them using Whisper, chunks & embeds them with Sentence Transformers, stores them in a vector DB (Pinecone), and en…

    Python

  3. Flourish Flourish Public

    Flourish is a smart plant care web application that bridges the gap between wanting to keep plants alive and actually knowing how. It uses real horticultural data combined with artificial intellige…

    TypeScript 1

  4. Entropy-Community-Forum Entropy-Community-Forum Public

    Entropy — A study ecospace and agentic study assistant: Built in Python with FastAPI, I developed async microservice endpoints that generate personalized quizzes and study content using pluggable L…

    TypeScript

  5. cross-portfolio-default-risk cross-portfolio-default-risk Public

    Implemented a supervised machine learning pipeline for loan default prediction using public loan-level data, inspired by prior ACM research. Performed leakage-aware preprocessing, feature engineeri…

    HTML

  6. socio-economic-crime-study socio-economic-crime-study Public

    Do neighborhoods located at sharp socio-economic boundaries experience higher crime rates than socio-economically homogeneous areas, even after controlling for poverty and spatial autocorrelation?

    Python 1