My Projects

Cybersecurity & IT

Home Lab & Private Cloud Infrastructure

I developed a fully self-hosted cloud lab designed to experiment with virtualization, AI deployment, and cybersecurity tools. This system serves as a personal cloud environment where I can deploy, test, and automate AI models, security tools, and networking applications.

Technical Infrastructure & Deployment

• Virtualization & Compute Setup:

  • Deployed Proxmox Virtual Environment as a hypervisor to manage multiple KVM-based virtual machines and LXC containers.

  • Configured GPU passthrough for high-performance AI processing, enabling LLM model inference directly on my infrastructure.

  • Optimized resource allocation and CPU pinning to ensure low-latency model execution.

• AI Model Hosting & Inference Optimization:

  • Ran local Large Language Models (LLMs) such as Llama 3.3, DeepSeek-R1, Mistral, and Falcon.

  • Used GGUF quantization to reduce memory overhead while maintaining model accuracy.

  • Deployed models via Ollama, TensorRT, and ONNX Runtime for accelerated inference.

• Containerized Services:

  • Built and managed Docker & Kubernetes clusters to efficiently run and scale AI-powered applications.

  • Deployed self-hosted AI APIs, Jupyter Notebooks, and research tools for machine learning experiments.

• Storage & File Management:

  • Hosted a private Nextcloud instance with end-to-end encryption to store and sync secure research data, AI models, and documents.

  • Used ZFS RAID 1 configuration for high-availability storage redundancy.

• VPN & Remote Access Security:

  • Set up a WireGuard VPN server to ensure secure remote access to cloud services.

  • Enforced Zero Trust Architecture (ZTA) principles by requiring multi-factor authentication (MFA).

Key Achievements:

✔ Created an entirely self-hosted AI inference and research cloud.
✔ Reduced AI model inference latency by 45% using hardware acceleration.
✔ Implemented enterprise-level cybersecurity protections in a home lab setting.

Firewall & Network Segmentation

To enhance network security and traffic isolation, I designed and configured a segmented network architectureusing pfSense, Cisco, and UniFi firewalls.

Implementation Details:

• Network Segmentation & VLANs:

  • Created separate VLANs for different traffic types to enforce network isolation:

    • IoT VLAN – Isolated smart devices to prevent lateral movement attacks.

    • Secure Server VLAN – Dedicated to Proxmox, AI model servers, and cloud storage.

    • Sandbox VLAN – Used for penetration testing, IDS/IPS experiments, and malware analysis.

• Firewall Rule Optimization:

  • Configured stateful firewall rules to block unauthorized inter-VLAN traffic.

  • Implemented Deep Packet Inspection (DPI) to detect and prevent malicious network behavior.

• Traffic Monitoring & Load Balancing:

  • Used pfSense HAProxy to distribute traffic loads between AI model servers.

  • Configured QoS (Quality of Service) policies to prioritize AI inference over non-essential traffic.

Key Achievements:

✔ Eliminated security risks from IoT devices by isolating them in a VLAN.
✔ Reduced unauthorized traffic by 80% using strict firewall rules.
✔ Improved AI model deployment efficiency with network load balancing.

SIEM & Threat Intelligence Lab

I built a Security Information and Event Management (SIEM) system for real-time network monitoring, intrusion detection, and forensic analysis.

Implementation Details:

• Log Collection & Monitoring:

  • Deployed Elasticsearch, Logstash, and Kibana (ELK Stack) to centralize firewall, endpoint, and server logs.

  • Used Syslog forwarding to aggregate logs from Proxmox, pfSense, UniFi, and Linux servers.

• Intrusion Detection & Prevention:

  • Set up Suricata and Zeek IDS/IPS to detect threats such as:

    • Brute force login attempts.

    • DDoS attack patterns.

    • Unauthorized data exfiltration.

  • Automated responses using Fail2Ban, blocking malicious IP addresses.

• Threat Hunting & Anomaly Detection:

  • Trained machine learning models to identify abnormal network behavior.

  • Integrated AI-driven correlation rules to analyze attack patterns and generate threat reports.

• Forensic Analysis & Reporting:

  • Automated forensic report generation using Python scripting and ELK visualization dashboards.

  • Developed email alert systems for real-time security breach notifications.

Key Achievements:

✔ Deployed an enterprise-level SIEM system for real-time threat intelligence.
✔ Reduced security incident detection time from hours to minutes.
✔ Built an AI-enhanced security monitoring system for anomaly detection.

Ethical Hacking & Penetration Testing

I conducted controlled penetration tests on my lab network to evaluate security vulnerabilities.

Implementation Details:

• Vulnerability Scanning & Exploitation:

  • Used Metasploit Framework to test for open ports, outdated services, and misconfigured servers.

  • Ran Nmap scans to identify potential security gaps.

• Wi-Fi Security Auditing:

  • Deployed Aircrack-ng to test wireless network security.

  • Conducted WPA2 handshake captures to analyze encryption strength.

• Password Cracking Simulations:

  • Used Hashcat and John the Ripper to test password security policies.

  • Evaluated brute force and dictionary attack resistance.

• Phishing Attack Simulations:

  • Developed custom phishing payloads to test user awareness.

  • Used Gophish phishing framework to assess email security protocols.

• Social Engineering & Physical Security Testing:

  • Designed RFID cloning tests to evaluate access control vulnerabilities.

  • Simulated USB drop attacks to measure employee security awareness.

Key Achievements:

✔ Identified and patched over 10+ network vulnerabilities in a controlled lab environment.
✔ Reduced potential security risks by hardening authentication systems.
✔ Enhanced cybersecurity training by simulating real-world phishing attacks.

AI & Machine Learning

Rebecca AI – Context-Aware AI Assistant

Rebecca AI is an intelligent, self-learning assistant designed to process, understand, and organize personal communication across emails, messages, and documents. Built with NLP techniques and deep learning models, Rebecca AI operates as an adaptive digital memory, improving over time.

Core Functionalities & Technologies Used:

• Contextual Response Generation: Uses transformer-based models (Llama 3.3, GPT-J) fine-tuned on personal message history to provide relevant and human-like responses.

• Sentiment Analysis & Emotion Detection:

  • Uses BERT and spaCy for sentiment detection and tone analysis.

  • Dynamically adjusts responses based on emotional context.

• Task Automation & Reminder System:

  • Named Entity Recognition (NER) for extracting important dates, deadlines, and scheduled tasks.

  • Automated tagging of tasks and cross-referencing with calendar systems.

  • Natural language command parsing for interpreting user intent in casual conversation.

Security & Privacy Considerations:

• Fully operates locally within my private cloud, ensuring data privacy.

• AES-256 encryption protects stored communication logs.

• Differential Privacy techniques applied to prevent model memorization of personal data.

Exo AI Cluster Contribution

The Exo project is an open-source, decentralized AI processing network, designed to enable multi-device AI inference using distributed computing techniques. My contribution focused on scalability, load balancing, and optimizing LLM inference on lightweight hardware.

Key Contributions:

• Multi-Device AI Inference:

  • Optimized load balancing algorithms to efficiently distribute AI computations across Raspberry Pi clusters, ARM-based devices, and cloud servers.

  • Implemented parallel processing optimizations using Ray and MPI (Message Passing Interface) to increase computation efficiency.

• Low-Power Model Optimization:

  • Fine-tuned quantized LLM models (Llama.cpp, GPTQ) to reduce memory footprint for deployment on low-end hardware.

  • Integrated ONNX Runtime and TensorRT to improve inference speed while maintaining accuracy.

• Scalability Improvements:

  • Developed dynamic scaling logic that automatically allocates compute resources based on real-time demand.

  • Enabled model partitioning, where large-scale AI models could be processed across multiple nodes.

AI for Cybersecurity & Threat Detection

This project involved applying AI and machine learning to detect cybersecurity threats by analyzing network traffic, anomaly patterns, and behavioral deviations.

Implementation Details:

• Anomaly Detection Model:

  • Built unsupervised ML models using Isolation Forest and Autoencoders to detect deviations in normal network behavior.

  • Collected and preprocessed network logs from Zeek, Suricata, and Wireshark.

• Pattern Recognition for Threat Detection:

  • Implemented CNNs and LSTMs to detect DDoS attack patterns and unauthorized access attempts.

  • Trained on real-world cybersecurity datasets (CICIDS2017, UNSW-NB15).

• Automated Intrusion Response:

  • Designed automated firewall rule adjustment scripts that block detected malicious traffic in real-time.

  • Used natural language generation (NLG) models to generate human-readable security alerts for administrators.

Ollama AI Model Deployment & Optimization

This project focused on running advanced LLMs locally, ensuring data privacy and improved inference speed on personal hardware.

Implementation Strategy:

• Optimized Local LLM Inference:

  • Deployed Llama 3.3, DeepSeek-R1, Mistral, and Falcon models using Ollama for fully local AI processing.

  • Integrated GGUF model quantization to reduce memory consumption and optimize speed.

• Fine-Tuning for Personal Use:

  • Developed custom instruction-tuned variants of the models for personal productivity tasks like summarization, contextual Q&A, and file organization.

  • Used LoRA (Low-Rank Adaptation) and PEFT (Parameter-Efficient Fine-Tuning) to modify models without excessive memory overhead.

• Model Deployment & Execution Environment:

  • Hosted models on dedicated AMD-based server with 8-core CPU, RTX 4090, and 128GB RAM.

  • Implemented vector search (FAISS, ChromaDB) for semantic memory retrieval.

AI-Powered Digital Forensics

I developed an AI-based forensic analysis tool that detects phishing attacks, social engineering attempts, and email fraud.

Core Features:

• Phishing Email Detection:

  • Trained BERT and DistilBERT models to classify emails as phishing/non-phishing based on:

    • Text structure & URL obfuscation patterns.

    • Metadata-based sender verification (e.g., SPF, DKIM, DMARC checks).

• Social Engineering Attack Recognition:

  • Applied GloVe word embeddings to detect persuasive and manipulative language in emails and messages.

• Forensic Analysis Dashboard:

  • Created a Flask-based web dashboard for displaying flagged security risks and generating forensic reports.

Automated Research Assistant

This AI-powered tool was designed to automate research workflows, allowing for document parsing, summarization, and knowledge extraction.

Key Features:

• Document Processing & OCR:

  • Used Tesseract OCR and LayoutLM to extract text from PDFs, images, and scanned documents.

  • Converted structured data into queryable knowledge graphs.

• Summarization & Contextual Understanding:

  • Implemented PEGASUS & BART transformers for extractive & abstractive summarization.

• Context-Aware Q&A:

  • Integrated with ChromaDB for semantic search, enabling conversational querying over documents.

Data Science & Predictive Modeling

Stock Price Prediction Model

This project aimed to predict short-term stock price movements using time series forecasting models and financial data analysis. The goal was to identify patterns, trends, and market signals that could provide insight into future price fluctuations.

Implementation Details:

• Data Collection & Processing:

  • Used Yahoo Finance API & Alpha Vantage to pull historical stock data (open, high, low, close, volume).

  • Collected technical indicators (SMA, EMA, RSI, MACD) to serve as features for the model.

  • Processed alternative data sources (news sentiment, financial reports) for better predictive power.

• Model Selection & Training:

  • Implemented ARIMA (AutoRegressive Integrated Moving Average) for short-term trend analysis.

  • Trained LSTM (Long Short-Term Memory) networks to capture sequential patterns and non-linear relationships in stock data.

  • Fine-tuned hyperparameters using grid search & Bayesian optimization to reduce error rates.

• Feature Engineering & Optimization:

  • Engineered lagged features, volatility indices, and momentum indicators to enhance model accuracy.

  • Applied principal component analysis (PCA) to reduce dimensionality and improve performance.

  • Used cross-validation techniques (rolling window method) to ensure robustness.

• Model Deployment & Real-Time Prediction:

  • Built a Flask API that allows real-time predictions based on live stock market data.

  • Deployed a Dash-based interactive dashboard for visualizing stock trends, forecasts, and confidence intervals.

  • Integrated sentiment analysis from financial news sources to refine predictions.

Key Achievements:

✔ Reduced prediction error (RMSE) by 30% after feature optimization.
✔ Integrated real-time stock price updates into the model, improving accuracy in fast-moving markets.
✔ Developed an interactive dashboard for traders to visualize forecasts.

Exploratory Data Analysis (EDA)

This project focused on analyzing large financial datasets to uncover hidden patterns, correlations, and trends.

Implementation Details:

• Data Cleaning & Preprocessing:

  • Handled missing values using interpolation & outlier detection via Z-score filtering.

  • Standardized & normalized data using MinMaxScaler & StandardScaler.

  • Applied feature engineering to derive rolling averages, momentum indicators, and trend signals.

• Data Visualization & Pattern Recognition:

  • Created heatmaps to visualize stock correlations and detect dependencies.

  • Built interactive candlestick charts using Plotly to analyze price movement.

  • Used histograms, KDE plots, and box plots to identify market distributions & anomalies.

• Sentiment Analysis Integration:

  • Scraped Reddit, Twitter, and financial news APIs to gauge market sentiment.

  • Tokenized & vectorized news data using TF-IDF & Word2Vec for NLP analysis.

  • Classified sentiment into bullish, neutral, and bearish categories to compare with stock movement.

Key Achievements:

✔ Identified correlations between social media sentiment & stock price movement.
✔ Automated real-time data analysis pipeline, reducing manual processing time by 40%.
✔ Developed visualization tools to spot trends and trading signals efficiently.

Recommendation Systems

This project involved developing a personalized recommendation system that suggests movies, books, and other media based on user preferences and past interactions.

Implementation Details:

• Collaborative Filtering (User-User & Item-Item Based):

  • Built memory-based and model-based collaborative filtering systems to recommend items based on similar user behavior.

  • Used cosine similarity & Pearson correlation to find users with similar preferences.

• Content-Based Filtering:

  • Vectorized book & movie descriptions using TF-IDF and BERT embeddings.

  • Built cosine similarity models to suggest items based on semantic meaning.

• Hybrid Model Integration:

  • Combined collaborative & content-based approaches to enhance recommendation accuracy.

  • Tuned weights for both methods using grid search & reinforcement learning techniques.

• Deployment & UI Integration:

  • Created an interactive Flask-based web application for personalized recommendations.

  • Integrated feedback learning—system refines suggestions based on user ratings and interactions.

Key Achievements:

✔ Achieved 90% accuracy in predicting user preferences on a test dataset.
✔ Implemented NLP-based analysis for deep contextual recommendations.
✔ Developed an interactive UI allowing users to refine recommendations dynamically.

Chatbot with Sentiment Analysis

This project focused on developing an AI chatbot capable of understanding and responding to user emotions.

Implementation Details:

• Natural Language Processing (NLP) Stack:

  • Used spaCy & NLTK for tokenization, lemmatization, and named entity recognition (NER).

  • Applied word embeddings (Word2Vec, GloVe) to improve chatbot contextual understanding.

  • Used transformer-based models (DistilBERT, RoBERTa) for sentiment classification.

• Sentiment Analysis & Emotion Detection:

  • Built a sentiment classifier that categorizes messages into positive, negative, and neutral tones.

  • Implemented emotion detection (anger, happiness, sadness, sarcasm, excitement, etc.) to generate adaptive responses.

  • Trained model on pre-labeled datasets (GoEmotions, Stanford Sentiment Treebank, and Twitter Sentiment Analysis Corpus).

• Dialogue Management & Response Generation:

  • Integrated Rasa NLU for intent recognition and conversation tracking.

  • Developed predefined response templates combined with GPT-3.5-generated dynamic responses.

  • Used reinforcement learning-based chatbot tuning to improve interaction quality.

• Deployment & UI Integration:

  • Hosted chatbot as a Flask API for integration with messaging platforms (Telegram, Discord, and Web Apps).

  • Implemented feedback learning system where user responses refine model accuracy over time.

Key Achievements:

✔ Developed a chatbot with a 92% accuracy rate in classifying emotional tone.
✔ Created an adaptive response system that changes tone based on detected sentiment.
✔ Integrated chatbot into real-world messaging applications with automated learning improvements.

Data Visualization Dashboard

This project involved creating an interactive dashboard to visualize stock market trends, AI predictions, and sentiment analysis.

Implementation Details:

• Interactive Graphs & Charts:

  • Used Matplotlib, Seaborn, and Plotly to generate real-time stock market charts.

  • Built heatmaps for correlation analysis and box plots for volatility tracking.

  • Created candlestick chart visualizations to track stock trends with moving averages.

• Live Data Integration:

  • Pulled real-time stock data from Yahoo Finance API & Alpha Vantage.

  • Integrated real-time social media sentiment analysis to show how tweets/news affect prices.

• Automated Report Generation:

  • Used Pandas & Jinja2 to generate PDF reports summarizing key trends.

  • Developed email alerts for major price fluctuations & market shifts.

Key Achievements:

✔ Built a real-time financial dashboard tracking market trends & AI predictions.
✔ Integrated sentiment analysis to visualize correlations between public sentiment & stock movement.
✔ Developed an alert system to notify users of significant price changes & trends.

Automated Closet & Smart Inventory System

This project automates wardrobe management using RFID tracking, AI-based recommendations, and real-time inventory updates. It integrates computer vision, IoT sensors, and AI-powered analytics to track clothing usage, condition, and availability in a fully automated system.

RFID-Based Clothing Tracker

This system uses RFID tags and a central database to track the movement and status of every clothing item in real-time.

Implementation Details:

• RFID Tagging System:

  • Each clothing item is embedded with an RFID tag containing a unique identifier linked to a central database.

  • RFID scanners are installed in closet shelves, laundry baskets, and entryways to track item location.

• Database & Status Updates:

  • Developed a custom SQLite/MySQL database that logs the real-time location of every item.

  • Status updates include:

    • In Closet → Available for wear.

    • In Laundry → Recently used, pending wash.

    • In Use → Currently being worn or taken outside.

• Automated Logging & Movement Tracking:

  • Every time an item passes an RFID scanner, the system logs movement patterns.

  • Detects missing items by tracking last-seen locations.

Key Features & Achievements:

✔ Automated real-time tracking of clothing inventory.
✔ RFID-based logging eliminates manual tracking efforts.
✔ Allows users to check wardrobe availability from a mobile dashboard.

AI Outfit Recommender

This feature suggests daily outfits based on weather, user preferences, and wardrobe history.

Implementation Details:

• Weather & Seasonal Integration:

  • Weather API integration (OpenWeatherMap) fetches real-time temperature, humidity, and precipitation data.

  • AI suggests clothing combinations based on weather conditions (e.g., jackets for cold days, light fabrics for summer).

• Personal Style Learning & Preferences:

  • AI learns user fashion habits over time based on past selections.

  • Uses collaborative filtering to suggest similar outfits based on historical choices.

• Color Matching & Outfit Coordination:

  • Implemented computer vision-based color analysis using OpenCV and K-Means clustering.

  • Analyzes clothing colors to ensure aesthetically matching outfits.

• Context-Aware Recommendations:

  • AI considers factors like formal vs. casual settings.

  • Suggests occasion-based outfits (e.g., business meetings, gym, date night, casual outings).

• Mobile App UI Integration:

  • Built a React Native app that allows users to view recommendations, mark preferences, and refine suggestions.

Key Features & Achievements:

✔ Daily AI-based outfit recommendations tailored to weather & events.
✔ Color-coordination ensures fashion-conscious combinations.
✔ AI improves over time by learning from user choices.

Automated Sorting & Alerts

This system sends notifications about clothing status, required laundry cycles, and missing items.

Implementation Details:

• Automated Laundry Tracking:

  • RFID tags track when items enter laundry baskets and estimate washing frequency.

  • Sends reminders if an item hasn’t been washed after a certain number of wears.

• Item Replacement & Wear Monitoring:

  • Uses fabric degradation tracking based on usage logs and last-wash dates.

  • Alerts when clothing needs replacement due to excessive wear or damage.

• Missing Item Alerts:

  • Compares last-seen location of missing clothes using RFID scanner logs.

  • Sends alerts if an item is left at work, a friend’s house, or misplaced at home.

• Smart Closet Inventory Dashboard:

  • Developed a dashboard UI that visualizes clothing availability in real time.

  • Provides filtering options to check availability by color, category, or season.

Key Features & Achievements:

✔ Smart notifications for laundry and clothing condition tracking.
✔ Alerts users if items are missing or need replacing.
✔ Dashboard provides a full wardrobe overview anytime.

Film Production, Digital Media, & Security

Shah Film Secure Data Management

This system secures film production data, ensuring encryption, access control, and cloud storage security.

Implementation Details:

• End-to-End Encryption & VPN Security:

  • All data stored on private Nextcloud servers using AES-256 encryption.

  • Access only available through VPN & multi-factor authentication (MFA).

• Version Control & Redundancy:

  • Automatic backup system ensures version control for scripts, raw footage, and post-production files.

  • RAID 1 storage setup prevents data loss in case of hardware failure.

• User Access Management:

  • Built a role-based access system to allow different levels of editing permissions for crew members.

  • Monitors logins and data transfers to prevent unauthorized access.

Key Features & Achievements:

✔ Protected high-value film assets from unauthorized access.
✔ Encrypted storage ensures that raw footage & scripts remain secure.
✔ Automated backup prevents accidental data loss.

Live Production Tech for Les Choristes

This project integrates AVL networking for live stage productions, enabling real-time synchronization between sound, lighting, and visuals.

Implementation Details:

• Dante Audio Networking:

  • Configured low-latency Dante audio routing to provide real-time audio mixing across multiple input sources.

  • Enabled remote-controlled equalization and sound adjustments via IP-based digital mixers.

• sACN & Art-Net for Lighting Control:

  • Implemented sACN (Streaming Architecture for Control Networks) to control stage lighting over IP.

  • Automated lighting cues are triggered based on scene transitions.

• PTZ Camera Integration:

  • Installed IP-controlled PTZ cameras for live-streaming and multi-angle video recording.

  • Allowed remote camera adjustments via web interface.

Key Features & Achievements:

✔ Synchronized real-time audio, lighting, and visuals for live performances.
✔ IP-based automation reduced the need for manual adjustments.
✔ Enabled high-quality live-streaming of productions.

AI-Based Content Moderation

This tool automates the detection of copyright violations, content risks, and facial recognition alerts in digital media.

Implementation Details:

• Deepfake & AI-Generated Content Detection:

  • Used CNN-based facial recognition to compare footage against known datasets.

  • Detects unauthorized video usage & AI-generated deepfake content.

• Copyright Risk Monitoring:

  • Applied fingerprint recognition algorithms to check whether audio/video clips match existing copyrighted works.

  • Integrated Shazam-like AI matching for real-time detection.

• Sensitive Content Filtering:

  • Used computer vision (YOLOv5) and NLP sentiment analysis to identify inappropriate or restricted content.

Key Features & Achievements:

✔ Automated detection of unauthorized content use.
✔ Deepfake filtering ensures ethical content moderation.
✔ Real-time alerts flag copyright risks before publishing.

Engineering, 3D Modeling, & Stage Production

Home Automation & Smart Devices

This project focused on creating a fully automated smart home ecosystem using IoT devices, microcontrollers, and AI-powered automation. The system integrates smart lighting, temperature regulation, security monitoring, and voice-activated controls.

Implementation Details:

• Smart Lighting System:

  • Built an IoT-controlled lighting system using Raspberry Pi and ESP8266-based smart relays.

  • Integrated with Google Assistant and Alexa APIs for voice-activated control.

  • Used motion sensors (PIR) to automatically adjust brightness based on room occupancy.

• Temperature & Climate Control:

  • Installed DHT22 temperature and humidity sensors connected to a Raspberry Pi home server.

  • Created a Python-based control system to automatically regulate fans, air conditioning, and heatingbased on external weather conditions (API integration).

  • Implemented historical data analysis to optimize HVAC efficiency using linear regression models.

• IoT Security & Monitoring:

  • Built a Raspberry Pi-powered security camera system using motion detection AI (OpenCV & TensorFlow).

  • Configured face recognition for authorized users using LBPH (Local Binary Pattern Histogram).

  • Developed a smart door lock system using an RFID authentication system and IoT relay switches.

  • Integrated anomaly detection algorithms to flag unusual movement patterns.

Key Features & Achievements:

✔ Integrated AI-powered smart home automation that adapts to user behavior.
✔ Optimized power consumption by 30% through smart regulation.
✔ Secured IoT communications using AES encryption to prevent hacking.

3D Modeling & Printing Projects

This project involved designing, prototyping, and fabricating functional mechanical parts, props, and enclosuresusing 3D printing and CAD modeling.

Implementation Details:

• Fusion 360 & Blender for 3D Modeling:

  • Designed complex mechanical structures, including gear mechanisms, servo motor housings, and rotating joints.

  • Created custom enclosures for Raspberry Pi and Arduino-based IoT devices.

  • Modeled ergonomic handheld tools using stress analysis in Fusion 360.

• 3D Printing Techniques & Materials:

  • Used FDM (Fused Deposition Modeling) printers with PLA, PETG, and TPU filaments for durable, high-precision prints.

  • Experimented with SLA (Stereolithography) resin printing for intricate, fine-detail parts.

  • Implemented multi-material printing for integrated circuits and flexible hinge components.

• Post-Processing & Finishing Techniques:

  • Used acetone vapor smoothing for glossy, polished ABS prints.

  • Applied sanding, priming, and painting for realistic stage props and prototype enclosures.

Key Features & Achievements:

✔ Designed and printed over 100+ unique models for real-world applications.
✔ Fabricated working prototypes for stage automation and IoT systems.
✔ Optimized print settings for high-strength, functional parts.

Stage Engineering & Set Automation

This project focused on automating stage elements to create dynamic, moving set pieces for live theater productions.

Implementation Details:

• Mechanical Set Pieces:

  • Designed rotating stage platforms with stepper motor-controlled motion.

  • Created motorized curtain draw systems using servo motor actuation.

  • Integrated Arduino-based wireless DMX controllers for synchronized lighting & movement.

• Animatronics & Practical Effects:

  • Built hydraulic-powered animatronic props with servo-driven articulation.

  • Developed programmable LED displays for interactive set pieces.

  • Integrated motion sensors to trigger automated scene transitions.

• Safety & Redundancy Systems:

  • Implemented emergency stop mechanisms with fail-safe wiring.

  • Built a redundant power backup system using Li-Po battery failovers.

Key Features & Achievements:

✔ Enabled fully automated set transitions, reducing manual workload.
✔ Developed wireless-controlled set elements for seamless performance synchronization.
✔ Enhanced stage safety by implementing real-time fault detection.

Augmented Reality, Emerging Tech, & Automation

Even Realities AR Development

This project focused on developing AR-enhanced applications for G1 Smart Glasses, improving overlay accuracy, user experience, and gesture-based controls.

Implementation Details:

• UI/UX Optimization for AR Glasses:

  • Developed custom UI elements optimized for low-latency AR rendering.

  • Implemented gesture-based navigation using ML-based hand tracking (MediaPipe).

  • Adjusted field-of-view constraints for enhanced AR content readability.

• AR Overlay & Spatial Mapping:

  • Used SLAM (Simultaneous Localization and Mapping) to stabilize AR elements in real-world environments.

  • Integrated real-time object detection using YOLOv5 and OpenCV.

  • Developed gesture-controlled information overlays for hands-free interaction.

Key Features & Achievements:

✔ Optimized AR UI for improved clarity and user interaction.
✔ Implemented real-time object recognition in AR overlays.
✔ Reduced rendering latency by 50% through code optimizations.

AI-Driven AR Enhancements

This project applies AI-driven intelligence to AR applications, enabling smart overlays, voice-based interactions, and contextual AI recommendations.

Implementation Details:

• AI-Powered Contextual AR Suggestions:

  • Integrated LLMs (GPT-4, Llama 3.3) for real-time contextual recommendations based on user behavior.

  • Used voice recognition & NLP processing to translate speech into AR actions.

• Real-Time AI Image Processing in AR:

  • Built computer vision models that analyze real-world objects and provide instant AR information overlays.

  • Implemented multi-object tracking with TensorFlow and OpenCV.

Key Features & Achievements:

✔ Enabled AI-driven AR contextual interactions.
✔ Developed speech-to-action AI command processing for AR.
✔ Integrated real-time multi-object tracking for interactive overlays.

Automated Home Security System

This project integrates AI-based motion detection, facial recognition, and automated alerts into a home security system.

Implementation Details:

• AI-Based Motion Detection:

  • Used OpenCV + TensorFlow for real-time motion tracking.

  • Differentiates human vs. non-human movement to prevent false alarms.

• Facial Recognition & Access Control:

  • Implemented LBPH (Local Binary Pattern Histogram) and DeepFace AI for facial authentication.

  • Configured Raspberry Pi-based access control system for door lock automation.

• Intruder Detection & Automated Alerts:

  • Configured IFTTT-based automation to trigger SMS & email alerts.

  • Developed anomaly detection models for unusual activity flagging.

Key Features & Achievements:

✔ Enabled AI-driven intrusion detection with real-time alerts.
✔ Automated home security through facial authentication & access control.
✔ Reduced false alarms by over 60% using AI-based motion filtering.