Portfolio

AI products and data-driven systems built for real-world use.

My portfolio brings together projects where I combined AI engineering, data systems, and product thinking to solve practical problems. Each case study is meant to show not only the technology behind the work, but also the decision-making, architecture, and value delivered.

AI products built with clear business value and technical depth
Production-oriented architectures across LLMs, APIs, and data systems
A focus on usability, reliability, and explainable outcomes

Featured AI ProductMyHandyAI

MyHandyAI

Product walkthrough

Product overview

Project snapshot

Problem

DIY users often struggle to turn scattered repair information into confident action.

Focus

Deliver context-aware repair guidance using multimodal AI and retrieval-backed workflows.

Stack

Computer Vision
LLMs
LangChain
Vector Search
FastAPI
React
AWS

Role

AI Engineer / Technical Lead, focused on product architecture, retrieval design, and intelligent workflow orchestration.

Overview

MyHandyAI is an intelligent DIY assistant designed to help users understand, diagnose, and complete repair tasks with clearer step-by-step guidance. The goal was to create an AI product that could bridge the gap between raw technical information and practical action for everyday users.

I helped shape the system architecture around computer vision, vector search, and LLM workflows so the assistant could interpret user inputs, retrieve relevant repair knowledge, and respond with grounded, usable instructions instead of generic answers. The product was built to feel helpful in real scenarios, not just as a demo of AI capabilities.

Enterprise IntelligenceClient Signal EQ

Client Signal EQ

Product walkthrough

Home screen

Project snapshot

Problem

Enterprise communication contains critical signals, but they are hard to extract consistently at scale.

Focus

Detect sentiment, escalation risk, and client behavior patterns from email workflows in a way teams can act on.

Stack

Email Analytics
Sentiment Analysis
Behavioral Signals
LLMs
LangChain
FastAPI
React
MongoDB Atlas
Qdrant

Role

AI Engineer / Technical Lead, focused on signal design, language intelligence, and scalable architecture for AI-driven analysis.

Overview

Client Signal EQ is an AI platform designed to analyze enterprise email communication and surface patterns that matter to account teams and business leaders. The goal was to move beyond basic message review and create a system that could detect sentiment, escalation risks, behavioral trends, and early client signals.

I worked on shaping the AI architecture around language analysis, retrieval-backed reasoning, and structured signal extraction so the platform could turn large volumes of communication into more actionable insights. The value of the product was in helping teams spot relationship risks and opportunities earlier, with a clearer view of client dynamics over time.

Forecasting PlatformAnalytics AI

Analytics AI

Analytics walkthrough

Forecasting dashboard

Project snapshot

Problem

Forecasting outputs are often technically correct but hard for teams to explore, question, and operationalize.

Focus

Combine predictive modeling with AI-guided exploration, simulation, and interactive decision support.

Stack

Forecasting
Agentic AI
SAS
Python
Gemini
Time-Series Modeling
Simulation
Optimization
Interactive Analytics

Role

AI Engineer / Analytics Builder, focused on forecasting workflows, agentic analysis, and translating model output into practical insight.

Overview

Analytics AI was built around the idea that advanced forecasting and decision support should be easier to explore, explain, and act on. The project focused on combining predictive analytics with agentic AI workflows so users could move from raw data and model outputs to more interactive insight generation.

I contributed to a solution that brought together forecasting, simulation, and AI-assisted analytics for electricity consumption scenarios. The system was designed not only to generate predictions, but also to help users interrogate the results, understand possible trends, and explore operational decisions through a more natural analytical experience.

NLP Analytics ProjectSentiment Analysis with Databricks

Sentiment Analysis with Databricks

This project focused on turning unstructured user feedback into usable product and behavioral insight. By collecting Google Play Store review data and processing it with NLP workflows, the goal was to understand how users felt, what patterns were emerging, and where meaningful segments could be identified.

I worked on a pipeline that combined data collection, sentiment modeling, clustering, and visualization so the results could be explored in a practical way rather than remaining purely technical outputs. The project connected Databricks-based analytics with downstream reporting in Power BI, making it easier to translate review data into decision-ready patterns.

Project snapshot

Problem

User reviews contain rich product signals, but they are hard to analyze consistently at scale without structured NLP and segmentation.

Focus

Extract sentiment and behavioral patterns from review data using transformer models, clustering, and analytics dashboards.

Stack

Databricks
Python
Web Scraping
Transformer Models
Sentiment Analysis
Clustering
SQL Server
Power BI
NLP

Role

Data Scientist / Analytics Builder, focused on NLP workflow design, clustering analysis, and insight visualization.

Energy Analytics SystemEnergy Consumption Smart Meters

Energy Consumption Smart Meters

Energy Consumption Smart Meters focused on turning large-scale smart meter data into actionable forecasting and operational insight. The project was centered on understanding consumption behavior, identifying meaningful patterns, and building models that could support better planning around electricity demand.

I contributed to a workflow that combined time-series analysis, AI-assisted exploration, and interactive analytics to make energy data easier to interpret and use. The strength of the project was in connecting technical forecasting work with a more practical decision-support layer, helping transform raw utility data into clearer signals for planning and analysis.

Project snapshot

Problem

Smart meter data is rich in value, but difficult to turn into clear forecasting and planning insight at scale.

Focus

Build forecasting and analytical workflows that help explain consumption behavior and support better energy planning.

Stack

Smart Meter Data
Time-Series Forecasting
Python
SAS
Gemini
Anomaly Detection
Interactive Analytics
Energy Modeling

Role

AI Engineer / Analytics Builder, focused on forecasting, energy data interpretation, and interactive insight design.

Healthcare AI AssistantMedical Chatbot

Medical Chatbot

This project focused on building a medical chatbot designed to interact directly with users in a more natural and helpful way. The goal was to create a conversational system that could learn from realistic doctor-patient exchanges and generate responses that felt coherent, relevant, and medically grounded.

I trained an LSTM-based conversational model in TensorFlow on roughly 100,000 doctor-patient conversations, and also fine-tuned T5 and GPT-2 models on the same dataset for comparison. This allowed the project to evaluate different conversational modeling approaches side by side, comparing how sequence models and transformer-based architectures performed on the same healthcare dialogue task.

Project snapshot

Problem

Healthcare conversations require models that can respond in a coherent and context-aware way while handling realistic doctor-patient dialogue patterns.

Focus

Build and compare multiple conversational AI approaches for user-facing medical dialogue generation.

Stack

LLMs
LSTM
TensorFlow
T5
GPT-2
NLP
Python
Conversational AI
Healthcare Data
Model Comparison

Role

AI Engineer / NLP Builder, focused on conversational model training, fine-tuning, and comparative evaluation across LSTM and transformer-based architectures.

BI & Data EngineeringOperational Dashboard Suite

Operational Dashboard Suite

This project represents a body of analytics and data engineering work built around stakeholder-facing dashboards for operations, sales, portfolio management, revenue monitoring, and cancellation analysis. The goal was to transform raw business data into clear reporting experiences that leadership and frontline teams could use to monitor performance and make faster decisions.

The dashboard suite included support operations reporting, sales versus target tracking, account bundle segmentation, revenue and exam monitoring by owner, and sales versus cancellations trend analysis. Across these dashboards, the recurring focus was building strong KPI views, ranked comparisons, interactive filters, and visual breakdowns that surfaced bottlenecks, concentration points, and areas needing follow-up.

Project snapshot

Problem

Business teams often have data available, but not in a form that makes performance, bottlenecks, and trends easy to act on.

Focus

Build stakeholder-ready dashboards that combine KPI design, segmentation, trend reporting, and operational visibility.

Stack

Power BI
KPI Reporting
SQL
Data Modeling
Operational Analytics
Sales Analytics
Segmentation
Revenue Reporting
Cancellation Analysis
Stakeholder Dashboards

Role

Data Analyst / Data Engineer, focused on KPI design, dashboard architecture, interactive reporting, and translating business data into practical decision-support tools.