name: Omar Madjitov
role: PhD Researcher & ML Engineer
location: Atlanta, Georgia 🇺🇸
education:
program: PhD Computer Science
institution: Georgia State University
research_focus:
- Natural Language Processing
- Large Language Models
- Text Mining
- Applied ML for Decision Systems
current_work:
- Financial downturn prediction from SEC 10-K filings
- Geospatial & transactional revenue analytics
- Teaching Python as Graduate Teaching Fellow
languages: [English, Russian]- 🔬 PhD Researcher focused on NLP, LLMs, and applied ML for decision systems
- 🧠 Building end-to-end ML pipelines that turn unstructured text into supervised learning problems
- 🌐 Also a full-stack engineer — I ship production apps with React, Node, PostgreSQL, and AWS
- 🎓 Graduate Teaching Fellow teaching Python to students from non-technical backgrounds
- 📫 Reach me at omadjitov1@icloud.com or through my portfolio: omarbuilds.dev
🧠 Machine Learning & NLP
💻 Frontend & Full-Stack
🗄️ Data & Infrastructure
|
Financial Downturn Prediction from Regulatory Filings End-to-end NLP pipeline that downloads 10-K filings from SEC EDGAR, extracts Item 1A (Risk Factors) via deterministic section parsing, and joins each filing to post-filing market outcomes. Produces an audit-ready supervised dataset labeling filings as high-risk when the stock experiences a ≥30% drawdown within 90 days — powering downstream LLM evaluation and predictive risk modeling.
|
Clustering + Seasonal Decomposition at Scale Built an analytics pipeline converting raw accounting exports into normalized relational tables. Recovered customer geolocation with multi-stage resolution (65% high-confidence coverage), validated spatial clustering via the Hopkins statistic, and derived 6 operational territories capturing 75% of revenue.
|
|
Reframing a Failed Segmentation Task Diagnosed a dataset–label mismatch in a medical imaging task and reframed semantic segmentation into bounding-box detection to preserve correctness. Built a reproducible training/inference pipeline with custom COCO↔YOLO converters. mAP@50 = 0.905 · IoU = 0.73 · 0.91 pixel accuracy
|
Production Full-Stack Application Full-stack job-application tracking platform with user-scoped CRUD, JWT authentication, protected routes, analytics dashboards, and external job-feed ingestion. Deployed with environment-based configuration.
|
|
Next.js App Router in Production Responsive corporate website built with Next.js App Router, mobile-first layout, optimized asset loading, and production email integration. Improved navigation clarity to support lead generation.
|
Community Q&A Platform Community-driven Q&A platform with authentication, private routes, voting, threaded discussions, search, sorting, and filtering. Handled Supabase auth flows, session persistence, and relational data modeling.
|
- 🔭 Researching NLP pipelines for financial risk extraction from SEC filings
- 🌱 Deepening my work in LLM evaluation and applied ML for decision systems
- 👨🏫 Teaching Python to non-technical students as a Graduate Teaching Fellow
- 💡 Open to collaborations at the intersection of ML research and production systems
Whether it's applied research, a production system, or a messy dataset nobody wants to touch — I'm interested.
⚡ Fun fact: when I'm not in a terminal, I'm cooking, running, or tinkering with IoT gadgets. ⚡