Skip to content
View Nivedita-Saha's full-sized avatar
11 followers · 10 following

Block or report Nivedita-Saha

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.

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
Nivedita-Saha/README.md

Hi 👋, I'm Nivedita Saha

Typing SVG

Building learned, controllable, and interpretable systems: LLM-agent simulations, scientific machine learning, and explainable models, with an emphasis on honest evaluation and reproducible engineering.

How to reach me

📧 Email: sahanivedita10@gmail.com  |  🌐 Portfolio: nivsaha.com  |  💼 LinkedIn: Nivedita Saha  |  💻 GitHub: @Nivedita-Saha

Skills

Languages

Python SQL JavaScript

ML & Deep Learning

PyTorch scikit-learn XGBoost U-Net Neural%20ODE GNN

Multi-agent & Simulation

LLM Agents Mesa RL Gymnasium Monte Carlo

LLM Tooling

Ollama Groq RAG Hugging Face

Trust & Explainability

SHAP LIME Uncertainty

Data, Cloud & Engineering

PostgreSQL BigQuery AWS Azure Docker Git pytest Streamlit Power BI

Contribution Streak

streak

Contributions Graph

activity graph

Contribution Snake

github snake

Featured Projects

async-agent-message-bus

async-agent-message-bus figure

Async multi-agent LLM simulation engine with structured message-passing between dozens of concurrent agents, demonstrated on a hybrid-threat disinformation scenario. A single-lexicon metric reported the false narrative as dead; a two-lexicon measure caught it mutating into a generalised distrust that saturated every agent. → View repository

theory-of-mind-reputation-engine

theory-of-mind-reputation-engine figure

A multi-agent commons simulation where each LLM agent tracks reputation and trust through an explicit, typed theory-of-mind layer. The engine isolates the exploiter, yet the commons still collapses — reputation without enforcement is not enough. → View repository

tumour-immune-twin

tumour-immune-twin figure

A controllable digital twin of tumour-immune dynamics: a neural surrogate of a mechanistic cancer model, steered by a reinforcement-learning controller toward a healthy equilibrium. → View repository

cardiac-mri-trustworthy

cardiac-mri-trustworthy figure

A trustworthy cardiac MRI pipeline, from segmentation to explainable, uncertainty-aware diagnosis, evaluated across scanner vendors. → View repository

cell-cell-gnn

cell-cell-gnn figure

A controlled study of graph neural networks for cell-type classification on breast-cancer tissue, with an honest finding: a feature-only baseline beats GCN and GAT under high label homophily. → View repository

semg-edge-ai

semg-edge-ai figure

Compressing a 1D-CNN for wearable sEMG gesture recognition through quantisation, pruning, and knowledge distillation, benchmarked on accuracy, model size, and inference latency. → View repository

Interested in research-engineer and machine-learning roles in multi-agent systems, simulation, and trustworthy AI.

👀 Profile Views

profile views

Pinned Loading

  1. affect-state-engine affect-state-engine Public

    Multi-agent negotiation simulation where each agent's emotional state is an enforced state machine that gates which actions it can take.

    Python

  2. async-agent-message-bus async-agent-message-bus Public

    Async multi-agent LLM simulation engine with structured message-passing, demonstrated on a hybrid-threat disinformation scenario

    Python

  3. llm-negotiation-sim llm-negotiation-sim Public

    Local, free-to-run multi-agent LLM negotiation simulator with evolving agent stance, reputation tracking, and structured JSON logging — built with Python, Ollama, and pytest.

    Python

  4. monte-carlo-negotiation-outcomes monte-carlo-negotiation-outcomes Public

    Monte Carlo analysis of a multi-agent LLM negotiation simulator: 50 randomised runs, agreement-rate and stance-convergence statistics, built on Python + Ollama (llama3.2:3b)

    Python

  5. theory-of-mind-reputation-engine theory-of-mind-reputation-engine Public

    Multi-agent commons simulation where LLM agents track reputation and trust through structured theory-of-mind

    Python

  6. wildlife-forest-abm wildlife-forest-abm Public

    Agent-based model (Mesa) simulating herbivore space use under varying forest management intensity, grounded in Grimsö Research Station's land-use research context.

    Python