Highway RL Agent
Training and evaluating a reinforcement learning agent for highway driving, with a focus on reward design, behavior analysis, and failure cases.
- Reinforcement Learning
- PPO
- Evaluation
STUDYBUILDEVALUATE
ObadaAlsehli
Software Engineer · MSc at Istanbul Technical University
I study methods, build intelligent systems, and evaluate how well they work.
I design and evaluate intelligent systems at the intersection of image processing, applied AI, and full-stack engineering.
01. SELECTED WORK
SELECTED WORK
Intelligent Document Management (IDMS)
A completed six-person capstone: an intelligent document-processing system with retrieval-augmented generation. My contribution: backend services, data processing, API design. The AI and retrieval components were built by teammates.
- Backend engineering
- RAG integration
Fashion Attribute Recognition
One image, four attribute heads: material, color, pattern and texture. DINOv2 and ResNet-50 compared, with a reported mean macro-F1 of 0.765 versus 0.649 on silver labels.
- Computer Vision
- Attribute Evaluation
Pose Squat Counter API
MediaPipe pose landmarks turned into repetition counts and simple depth and torso-lean indicators for uploaded side-view MP4s. Not a clinical assessment.
- Pose Estimation
- API
02. RESEARCH & WRITING
RESEARCH & WRITING
Engineering deep dive
Liveness Is Not Model Readiness
What separate liveness, health, and readiness responses mean in a documented vision-model API, and what they do not establish about prediction quality.
Experiment report
Silver Labels Are Not Ground Truth: A Four-Head Fashion Classification Project
What do multi-head vision benchmarks measure when annotations come from heuristics and VLMs? This write-up examines the four-head classification setup, reported F1 metrics, and a separate 210-image human label audit.
Failure analysis
Degenerate Policy in a Highway RL Agent
A PPO agent learned to favor slow driving rather than the intended behavior. This write-up examines the reported training results and reward setup.
03. ABOUT
ABOUT
Research-minded engineer connecting experimental rigor with reliable software.
I document what I build, measure and learn.
Full backgroundI’m a software engineer and MSc student in Information & Communications Engineering at Istanbul Technical University, with a BSc in Software Engineering from Bahçeşehir University.
I enjoy turning technical questions into practical systems, combining research, engineering, and evaluation.
My contribution spans Python backend services, React application workflows, relational data modeling and document-processing integrations, alongside testing, technical documentation and handover preparation.
CORE AREAS
Computer VisionImage understanding and analysis
AI EvaluationExperimental rigor and error analysis
Backend & SystemsApplications and data pipelines
TECHNOLOGIES