// ANOMALY FLAGGED AT t=0.68 — see below

Building AI systems that catch
what shouldn't be there.

I'm Shivani Tiwari — an AI Engineer and Data Scientist working across machine learning, deep learning, and Generative AI. I build end-to-end ML pipelines, time-series anomaly detection systems, and RAG applications powered by large language models.

M.Tech (AI) · IIT Jodhpur GATE 2024 Qualified Salempur, Deoria, UP
View projects → Get in touch GitHub
scroll
01 · Signal Profile

About

The short version of the training run.

I'm an AI Engineer and Data Scientist currently finishing my M.Tech in Artificial Intelligence at IIT Jodhpur. My work sits at the intersection of applied deep learning and generative AI — from fairness-aware diagnostic models to LLM-powered retrieval systems.


I care about building models that hold up outside the notebook: pipelines that are reproducible, systems that are fast to deploy, and predictions that are fair and explainable. My toolkit runs from PyTorch and HuggingFace to LangChain, FastAPI, and AWS — with a habit of pruning, tuning, and stress-testing everything before calling it done.

5+
ML / GenAI Projects
23
Crypto Assets Analyzed
0.23s
RAG Response Latency
+38%
Fraud Recall Gain
02 · Event Log

Experience

Where the signal was collected.

AR/VR Intern
National Institute of Electronics and Information Technology (NIELIT), Tezpur
March 2024 – May 2024
  • Developed interactive data-driven AR learning applications using Unity3D and Vuforia for educational visualization.
  • Implemented a Blood Alcohol Concentration (BAC) analytics module simulating physiological effects from real-time user input.
  • Optimized rendering and system workflows for stable performance across multiple Android devices.
  • Conducted testing and validation to ensure reliable application deployment.
Research Intern
Centre for Development of Advanced Computing (C-DAC), Pune
July 2023 – August 2023
  • Evaluated network performance using Open5GS, analyzing streaming workloads across multiple video qualities.
  • Performed data-driven analysis of throughput, bandwidth, and latency metrics.
  • Generated insights to improve network efficiency and scalability.
03 · Detections

Projects

Five systems, five different kinds of anomaly, bias, and signal to catch.

Fairness-Aware Dermatological Diagnosis
repo →
  • Fairness-aware CNN for dermatology classification on the Fitzpatrick17K dataset.
  • SNNL-based filter pruning on VGG-11 to remove bias-inducing filters.
  • Fairness–accuracy trade-off evaluated with Equalized Odds and FATE.
M.Tech Thesis · bias-aware pruning
RAG System for Intelligent QA
repo →
  • End-to-end GenAI QA system using LangChain, Llama3, and ObjectBox vector database.
  • Semantic retrieval with HuggingFace BGE embeddings.
  • Streamlit interface for document ingestion and contextual querying.
0.23s average response latency
Crypto Anomaly Detection (LSTM + Transformers)
repo →
  • Multivariate time-series anomaly detection across financial assets.
  • LSTM Autoencoder and Transformer architectures compared head-to-head.
  • Optuna-based hyperparameter tuning to sharpen detection accuracy.
23 crypto assets, incl. Bitcoin
Credit Card Fraud Detection
repo →
  • Supervised ML pipeline built for a heavily imbalanced dataset.
  • SMOTE and SMOTEENN resampling to lift minority-class recall.
  • SHAP explainability integrated for transparent decisions.
+38% minority-class recall
Customer Segmentation (Unsupervised)
repo →
  • K-Means, GMM, DBSCAN, and Hierarchical clustering compared.
  • RFM-based feature engineering for sharper segments.
  • Clusters evaluated with Silhouette Score and Davies–Bouldin Index.
4 clustering methods benchmarked
04 · Stack Trace

Technical Skills

The instruments, tuned and ready.

Programming
PythonSQLC / C++
Data Science & Analytics
PandasNumPyScikit-learn MatplotlibSeabornPlotlySciPy
Machine Learning / Deep Learning
PyTorchTensorFlowKeras CNNsLSTMsTransformers AutoencodersAttention MechanismsTransfer Learning
Generative AI / LLMs
LangChainHuggingFace TransformersRAG Pipelines Vector DatabasesEmbedding ModelsPrompt Engineering LLM Fine-TuningPEFTLoRA LlamaMistral
Data Engineering & MLOps
Feature EngineeringModel OptimizationHyperparameter Tuning W&BRay TuneDocker AWS S3 / EC2 / SageMakerCI/CDModel Serving
API & Deployment
FastAPIStreamlitREST APIs ONNX RuntimeModel Quantization
Tools
GitJupyterVS CodeLinux
05 · Training Data

Education

The foundation model.

DegreeInstituteScoreYear
M.Tech, Artificial Intelligence IIT Jodhpur 7.27 CGPA 2024–2026
B.Tech, Computer Science & Engineering Tezpur University, Assam 7.98 CGPA 2020–2024
Senior Secondary CBSE Board 90.4% 2019
Secondary CBSE Board 10 CGPA 2017
Key courses — Data Structures, Algorithms, DBMS, Machine Learning, Deep Learning, Computer Vision, NLP, Artificial Intelligence, Large Language Models.
06 · Verified

Achievements

  • Qualified GATE 2024 — Computer Science
  • Introduction to Machine Learning — Acmegrade
  • Joy of Computing Using Python — NPTEL