Emre mrkckpln

Software Engineer & Focused on AI Automation

Architecting secure, high-performance, and data-driven intelligent systems across Software 2.0 and MLOps.

Rotakur: Unified Mobility & Fleet SaaS
Graduation Project — Enterprise B2B SaaS + B2C Marketplace
A resilient, multi-tenant hybrid system orchestrating corporate fleet operations, organization event logistics, and public mobility bookings.
- AI Dispatch & Spatial Routing: Deployed an NLP Smart Dispatcher (Gemini 2.5 Flash + pgvector) to extract structured reservation profiles from raw chat text. Coupled with OSRM multi-stop road-network polyline decoding, dynamic geofencing, and strict POI constraints (airports/terminals) to eliminate phantom bookings.
- Dual-Tenant Governance & P&L Engine: Governs multi-entity corporate procurement via a locked state machine (DRAFT -> APPROVED) with strict data isolation. Integrated an AI Surge Pricing engine calculating real-time binding fares and exact Net P&L margins (fuel + depreciation) prior to trip onset.
- Sustainability & Regulatory Compliance: Built a smart carpooling matching engine pairing B2C orders (within ±2h window / 20km radius) to track DEFRA-compliant CO2 savings for ESG reporting. Automates U-ETDS reporting for the Turkish Ministry of Transport and executes UBL-TR 2.1 e-invoicing via DLQ-safe fallbacks.
Stack: Python, FastAPI, SQLAlchemy 2.0, PostgreSQL, pgvector, WebSockets, Redis, OSRM, Docker

👁️ Autonomous AI Sentry (AI Bekçi)
Edge CV · Millisecond Anomaly & Environmental Variance Detection
An intelligent autonomous perception engine designed to act as an unblinking AI Sentry in critical environments.
- Sentry Architecture: Deployed YOLOv11-OBB and YOLO-Pose models to infer behavioral anomalies, falls, and environmental variations within milliseconds.
- HPC Optimization: Fully accelerated via TensorRT FP16 quantization and CUDA execution layers to minimize edge latency and maximize hardware throughput.
Stack: C++, Python, PyTorch, TensorRT, CUDA, FastAPI, Next.js

♟️ Intelligent Chess Workstation
Asynchronous Multi-Tenant Micro-SaaS
AI-driven analytics platform bridging brute-force engine data with abstract human positional understanding.
- Async Core: Engineered a decoupled Redis Queue (RQ) pipeline for per-ply Stockfish and LLM analysis.
- Advanced Reports: Generated grandmaster-level heuristic feedback alongside publication-quality PDFs.
Stack: Python, FastAPI, Next.js 15, Redis, Stockfish, OpenAI/Gemini API, fpdf2

⚖️ LegalLens — AI Contract Analyzer
Neural Network Chrome Extension
Detects high-risk clauses using a custom Bi-LSTM + Attention model and Google Gemini for deep summary.
Stack: Python, PyTorch, FastAPI, Chrome API

Distributed Systems & Multi-Tenant SaaS: Engineering asynchronous, high-concurrency backend infrastructures with FastAPI, SQLAlchemy 2.0, and robust tenant-isolated PostgreSQL databases.
Production-Grade AI & Edge CV: Constructing end-to-end computer vision pipelines (YOLO-OBB/Pose) and custom sequential models structured for real-time inference in critical environments.
Inference Optimization & HPC: Optimizing deep learning models via TensorRT FP16 quantization, minimizing latency, and maximizing hardware throughput utilizing CUDA acceleration.
Spatial Computing & Telemetry: Designing custom multi-stop routing logic, geographic validation systems (OSRM), and live operational telemetry pipelines via persistent WebSockets in Rotakur.

Agentic Engineering: Architecting strict system guidelines and advanced prompt workflows to orchestrate AI-assisted development across large-scale production codebases.
Strategic System Foresight: Translating professional-level chess foresight (15+ years, 2000 FIDE) into software engineering — anticipating architectural edge cases, race conditions, and system bottlenecks before they occur.
High-Performance Architectures: Deepening expertise in hardware-level performance tuning, Linux-native optimizations, and high-throughput data processing layers using Modern C++.