AI & Data Engineering student @ Fırat University · Bursa, Turkey
🎯 My focus: turning LLMs from chat toys into verifiable, autonomous research systems — where the model proposes and a deterministic engine proves.
- 🎓 AI & Data Engineering student at Fırat University
- 🤖 Building agentic AI systems — multi-agent loops with LLM planners and deterministic verifiers
- 📈 Deep interest in quantitative finance: backtesting, leakage control, statistical validation
- 🛰️ Working on a TEKNOFEST Electronic Warfare system — SDR signal processing + ONNX modulation recognition
- 🧪 Comfortable across the stack: Python · Django · Flask · PyTorch · LangChain · SQL · Docker
- 💡 I care about systems that are reproducible and honest — every experiment seeded, hashed and auditable
- 📫 Reach me at o.bugra16@gmail.com
Languages
AI / Machine Learning
Backend & Data
Tooling
|
An autonomous quant researcher. A closed loop: the LLM proposes a hypothesis → a safe DSL compiles it into a strategy → leak-free walk-forward backtest → statistical gate (Deflated Sharpe, PSR, bootstrap CI, BH-FDR) → accept/reject → memory → next hypothesis. The rule: the LLM never touches data or backtests. It writes no free Python — only a validated DSL.
|
A TEKNOFEST Electronic Warfare system. Streams live IQ samples from an SDR over ZMQ/GNU Radio, runs ONNX modulation recognition, and drives direction-finding, jamming and GPS/radio spoofing panels from a PyQt6 cockpit. Highlights: TDoA/AoA solvers, waterfall plots, and a graceful fallback when no model or SDR is attached.
|
|
Drop in a CSV/XLSX and an LLM drafts an analysis plan. You approve it — then the agent runs EDA, feature work and modelling, and writes the report itself. Human-in-the-loop by design.
|
A full Django REST Framework API over Borsa İstanbul data: RSI/MACD/EMA technical signals, Swagger docs, and a cache benchmark that generates up to 1M synthetic rows to prove HIT vs MISS latency.
|
|
A RAG chatbot for financial terminology — Google Gemini + LangChain + FAISS behind a Flask app, deployed on Render.
|
A crypto portfolio tracker on Flask + MSSQL, pulling live prices from the CoinGecko API and computing realised / unrealised P&L.
|
- Pushing
autoresearch_quanttoward a fully autonomous research campaign on real market data - Preparing the electronic warfare system for TEKNOFEST
- Reading up on multiple-testing correction and deflated performance metrics in quant research
- Open to internships and collaborations in AI / ML engineering and data-driven fintech
"The LLM proposes. The engine proves."