Junior Data Analyst · Computing Science grad (TRU, April 2026) — Victoria, BC, Canada 🇨🇦

I build data and ML systems end-to-end: Power BI / DAX / SQL for analytics, and Python for everything from fine-tuned LLMs to multi-agent automation pipelines. I care about the part most projects skip — making something work for someone else, not just on my own notebook.

🔭 Recent work: NovaMart — an end-to-end e-commerce analytics project (166K+ records, Snowflake → Power BI) that quantified a $2.01M operations-improvement opportunity with statistically validated findings.

💼 Open to work — Data Analyst · BI Analyst · Analytics Engineer roles — I turn messy data into decisions with SQL, Power BI/DAX, and Python, and build ML when the problem needs it. (On-site · Hybrid · Remote)

Analytics / BI: Power BI · DAX · SQL (Snowflake, Oracle) · ETL · Medallion / dimensional modeling · statistical testing (Mann–Whitney U, Chi-Square) · pandas · NumPy · Plotly

Languages: Python · C# · C++ · Java · JavaScript · PHP · SQL

ML / AI: PyTorch · TensorFlow / Keras · scikit-learn · Hugging Face · Mistral-7B (QLoRA/LoRA) · CUDA · DCGAN · LSTM · TF-IDF · OpenCV

LLM / Agents: Anthropic Claude (Opus/Sonnet/Haiku) · OpenAI GPT-5.5 · LangChain · LangGraph · n8n · RAG (pgvector)

Backend / Web: Django REST Framework · FastAPI · Flask · Supabase · PostgreSQL · Docker

Cloud / Data Platforms: AWS · Azure · Snowflake · Supabase

GIS: Folium · geographic information systems

• novamart-analytics — End-to-end e-commerce operations analysis: 166K+ records traced through a Medallion (Bronze→Silver→Gold) Snowflake pipeline, bottlenecks proven with statistical hypothesis testing, and a $2.01M improvement opportunity surfaced in a 5-page interactive Power BI dashboard.
• timeseries_forecasting — Multi-Model Stock Forecasting with Monte Carlo Risk Analysis — benchmarks regressors and classifiers on S&P 500 returns, statistically testing whether any signal beats an Efficient-Market baseline. Ships with an FVG paper-trading bot (Alpaca) and an AWS Lambda S3 backup pipeline.

• RAG Platform (TRU Capstone) — RAG platform using Django + Supabase + pgvector + OpenAI.
• llm_classifier_interface — News classifier: TF-IDF shortlister + fine-tuned Mistral-7B (QLoRA) inference, served via a Django REST API with a web UI. The hybrid serving stack cut inference cost while holding accuracy (0.82 macro F1, beating a Random Forest baseline).
• AI-Powered Document Automation (n8n) — Multi-agent pipeline orchestrating Claude + GPT-4o mini to turn a job description into a tailored, fact-checked resume and cover letter, with a deterministic "lie detector" audit layer and full application tracking.
• mousedynamics / keyboarddynamics — Behavioral-biometric bot detection from mouse and keystroke dynamics, with leakage-free participant-level evaluation.
• vladlen — TensorFlow/Keras DCGAN turning Soviet-era propaganda posters into 128×128 abstract artworks.
• nlptextcleaning — Document-level semantic news classification with noisy-label auditing.

• ToolApp (Workforce & Equipment Tracking Dashboard) — Flask GIS dashboard tracking tools, employees, and work zones on a Folium map with Plotly usage reports (Azure SQL).
• ToolFool — Equipment-renting system visualizing geographic locations and tracking tool usage history, deployed to Azure App Service.

• socialnetworkBackend — Subscription-based social network backend (Django REST Framework, PostgreSQL, social OAuth, AWS, Docker).
• webCopyAI — Converts a static HTML/CSS/JS site into an installable Joomla 4 template, via CLI or a Django web UI.

• Thompson Rivers University — B.Sc. Computing Science (Sep 2021 – Apr 2026). Capstone: RAG platform with Django + Supabase + pgvector + OpenAI. Coursework: Data Mining & BI, Machine Learning, Applied AI, Database Systems, Behavioral Biometrics, Software Architecture, IoT.
• Turan College — Diploma in Computer Technology & Software (graduated with honors).

• Fundamentals of Building AI Agents — Coursera (Jun 2026) — LangChain, LangGraph

English (full professional) · Russian (native/bilingual)

• LinkedIn: karennurlybekov
• GitHub: @knurlybekov
• Email: nurlybekov.karen@gmail.com