Retrieval
Search and memory systems that help software bring back the right context at the right time.
Smart Data Foundations. Purposeful AI System.
I build systems that make raw information easier to search, understand, and act on. My work is mostly around retrieval, local AI, text processing, analytics dashboards, and data workflows. The macOS apps are side projects where I explore product ideas and interface craft.
Generative AI
LLMs
SLMs
Natural Language Processing
Text Processing
Text Mining
Data Science
Machine Learning
Data Analytics
Prediction
AI systems
AI Agents
Retrieval-Augmented Generation
Prompt Engineering
Context Engineering
Deep Learning
Fine-Tuning
Vision
Diffusion Models
What I work on
AI and data projects that turn messy inputs into useful outputs.
Text and NLP
Projects around classification, redaction, unredaction, document cleanup, and structured extraction.
Analytics
Dashboards and pipelines that make patterns visible without forcing people to dig through raw files.
Selected work
AI, data, ML, NLP, macOS, and dashboard projects from my public GitHub.
AI / Personal data
FinAgent
A personal finance assistant that turns bank activity into a conversational workflow. FinAgent connects through Telegram, reads spending and balance data, and lets someone ask natural questions instead of manually sorting through transactions across multiple apps.
Focus: private, practical answers for everyday financial questions, with local-first AI when privacy matters and hosted models when speed matters.
View repositoryData / Research
EpigCorpus
A reproducible research pipeline for Latin epigraphy. It pulls inscription data into a cleaner working corpus, prepares normalized text variants, and makes the material easier to inspect through exports and an interactive Roman Empire map.
Focus: turn fragmented historical records into a structured corpus that can be searched, mapped, validated, and reused for analysis.
View repositoryML / Classification
Cuisine Predictor
A machine learning classification project built around ingredient lists. The workflow treats recipes as text-like feature sets, transforms ingredients into model-ready signals, and predicts the cuisine category from those patterns.
Focus: show the full ML loop clearly, from feature preparation and training to evaluating how well ingredient signals explain cuisine labels.
View repositorySide / macOS
Mac Apps
A collection of native macOS product experiments. MacMail explores Gmail and Calendar workflows, MacUtils gathers desktop helpers, Pomo Duck turns focus sessions into a playful utility, and QTI Converter handles assessment-format conversion for education workflows.
Focus: fast native workflows, thoughtful desktop interactions, and small tools that feel sharp enough to use every day.
ML / Text modeling
The Unredactor
A text modeling project that studies whether surrounding context can recover redacted names. It cleans the corpus, converts text into features, trains predictive models, and compares results with accuracy, precision, recall, and F1.
Focus: make the modeling process transparent, from data cleanup and feature engineering to measurable prediction quality.
View repositoryNLP / Redaction
The Redactor
The practical counterpart to the unredaction project: a text-processing utility for detecting sensitive content and producing safer versions of documents. It focuses on privacy, cleanup, and making text easier to share without exposing details that should stay hidden.
Focus: identify sensitive language, redact it consistently, and support safer document handling for downstream workflows.
Find repositoryData / Dashboards
Analytics Dashboards
A group of BI and spreadsheet dashboard projects covering bank loans, UK road accidents, and Netflix catalog data. Each project turns raw records into KPI surfaces with filters, trend views, summary cards, and visual breakdowns.
Focus: make dense datasets scannable, comparable, and useful for decisions through clear metrics and interactive reporting surfaces.
Get in touch
Interested in practical AI or data work?
I am most interested in projects where AI and data need to become something usable: search, extraction, dashboards, document workflows, classification, or local assistants.