Contact Me
Nikhil Dhiman - AI / ML Engineer

EvacAI

AI-Powered Indoor Evacuation Assistant with Real-Time Threat Adaptation


  • Developed an intelligent indoor evacuation assistant that dynamically guides users to safety by processing natural language reports of blocked or dangerous paths in real time.
  • Integrated a CrewAI + LangChain agent system to simulate a multi-step evacuation planner, using memory, tools, and feedback to generate adaptive escape routes based on live user input.
  • Implemented route updates via API calls triggered by user-reported threats, enabling rerouting and real-time communication with natural, human-like instructions using VAPI.

Tech Stack: Python, Flask, Supabase, OpenAI API, CrewAI, LangChain, VAPI








Artwork Mapped Using ML

An interactive 3D visualization of 120K artworks mapped by visual similarity using ML and dimensionality reduction.


  • Engineered a machine learning pipeline using ResNet50 to extract high-dimensional features from 120K artwork images, enabling visual similarity analysis across diverse art styles.
  • Applied PCA and UMAP to compress embeddings into 3D space, enabling structure-aware visualization of visual similarity across artworks.
  • Built an interactive 3D visualization interface, allowing intuitive exploration of artistic relationships and clusters.

Tech Stack: Python, Jupyter Notebook, Pandas, NumPy, Matplotlib, Seaborn, Scikit-learn, Plotly
Dataset: WikiArt Dataset








Early Skin Cancer Detection Using AI

A CNN-based skin cancer classification model trained on dermoscopic images to detect malignancies with high accuracy.


  • Implemented and fine-tuned an EfficientNetB0 model using TensorFlow/Keras for skin lesion classification, achieving 85%+ validation accuracy on dermoscopic image data.
  • Improved model generalization with effective preprocessing techniques such as image resizing, normalization, and data augmentation.
  • Streamlined training through early stopping and real-time metric visualization, reducing overfitting and ensuring efficient convergence.

Tech Stack: Python, Jupyter Notebook, TensorFlow, Keras, NumPy, Pandas, Matplotlib, Seaborn, EfficientNet (via Keras Applications)
Dataset: ISIC 2024 - Skin Cancer Detection with 3D-TBP








Amazon Employee Access Challenge

A multi-model machine learning system leveraging XGBoost, SVM, and ensemble learning to predict employee access rights with 90%+ accuracy

  • Built and evaluated a diverse set of eight machine learning models, including XGBoost, SVM, and ensemble-based Voting Classifiers, to predict employee access rights with over 90% accuracy on internal access log data.
  • Engineered and encoded categorical features using label encoding, and ensured complete data integrity with zero missing values.
  • Compared performance across models using classification reports (precision, recall, F1-score), identifying the most reliable algorithms for high-stakes access prediction.

Tech Stack: Python, Jupyter Notebook, Pandas, NumPy, Scikit-learn, Matplotlib, Seaborn
Dataset: Amazon Employee Access Dataset (Kaggle)








IMDb Movie Analysis

An exploratory analysis of IMDb movie data to uncover insights into profitability, genre trends, and director performance.


  • Performed in-depth exploratory data analysis on IMDb movie data to identify patterns in budget, revenue, genres, and directorial impact on movie success.
  • Created new profitability metrics by deriving and visualizing the difference between budget and gross revenue across hundreds of films.
  • Visualized trends using bar plots, histograms, and correlation heatmaps to uncover relationships between runtime, rating, and revenue.

Tech Stack: Python, Jupyter Notebook, Pandas, NumPy, Matplotlib, Seaborn
Dataset: IMDb Dataset








KNN Classification using Scikit-learn

To classify flower species using the Iris dataset via the K-Nearest Neighbors (KNN) algorithm and evaluate model performance.


  • Implemented and tuned a K-Nearest Neighbors (KNN) classifier on the Iris dataset, achieving 97.8% accuracy at k = 3.
  • Analyzed feature relationships through visualizations and correlation heatmaps to understand class separability in flower species.
  • Evaluated model performance across multiple values of k, using accuracy plots to select the optimal hyperparameter for classification.

Tech Stack: Python, Scikit-learn, Pandas, NumPy, Matplotlib, Jupyter Notebook
Dataset: Iris Dataset (UCI ML repository)
Tool

My AI/ML Expertise Includes, but Is Not Limited To

Feature Engineering

Expertise in cleaning, transforming, and engineering features from raw datasets to enhance model accuracy and reliability.

Machine Learning Modeling

Experienced in training and optimizing models using Scikit-learn, Pytorch, and TensorFlow, with a strong focus on model performance and generalization.

Model Evaluation & Experimentation

Proficient in using cross-validation, ROC-AUC, precision/recall, and custom metrics to rigorously test and compare ML models.

AI Infrastructure & Deployment

Skilled in deploying ML models using Docker, FastAPI, and cloud platforms (AWS/GCP) for real-time inference and scalable serving.

My Reviews

What people say about
working with me

Dr. Armando Beltran

Assistant Professor of Computer Science

Nikhil has consistently distinguished himself through his exceptional programming skills, critical thinking, and advanced mathematical knowledge, which are essential for solving complex problems in artificial intelligence. His academic performance in these courses has been outstanding, and based on his achievements, I would rank him in the top 1% of all students I have taught at Cal State LA.

Chandrapal Singh

Director & Co-Founder

I’m happy to working with Nikhil, he is having a good problem solving skill. I recommend him and his team.

Abichal Jha

iOS Developer @ Gartner

Nikhil is one of the best among all the people I have ever worked with. As I remember,he was a very productive person, hardworking, broad-minded and forward thinking individual. Intelligent, ambitious, energetic and proactive perfectionist. Desire for proficiency and education makes him a valuable asset to the team. Working with him is a signature of success.

Upendra Kumar Tiwari

Data Scientist | Machine Learning Engineer

You have been the hard working and sincere student of mine. whichever assignment was given to you, you completed them within time. The characteristic that you posses is your hardwork, always trying to learn new technology and your simplicity will always help to reach your goal.

Michael Hassey

Chief Technical Officer

I had the pleasure of working with Nikhil on the MPC project, where his technical expertise, proactive problem-solving, and collaborative mindset were truly remarkable. He consistently demonstrated a deep understanding of complex systems, offering innovative solutions that drove the project's success. Nikhil's ability to communicate effectively and his dedication to delivering high-quality results made him an invaluable team member. It was an absolute privilege to work alongside someone so skilled and committed, and I highly recommend him for any future opportunities.

Counters

Driven by passion, defined by results

A results-driven engineer with 4+ years of experience turning complex problems into simple, impactful solutions across software, mobile, data, and AI.

150+

Global collaboration across roles and domains, fueled by cross-functional teamwork.

10+

Awards & recognitions for innovation, dedication, and impact acros projects and teams.

12+

Successfully completed projects spanning multiple domains and industries

8+

Client and Employer reviews reflecting trust, satisfaction, and success

Contact Me

Got a question or an idea? Feel free to reach out! Fill out the form below, and I’ll get back to you soon.

  • United States of America (USA)

  • hello@nikhildhiman.me