Software Developer · Applied AI & Computer Vision
Email: m.kadri.ensari@gmail.com
About Me / Bio
- Software developer and AI engineer with 4+ years building production systems — full-stack web and mobile, computer vision, LLM automation, and applied deep learning for commercial and research use cases.
- Delivered 15+ live products: real-time fleet telemetry, an interactive 3D vehicle configurator showcased at a Germany industry event, underwater sonar pipelines (YOLOv8 + SAM), AI-driven CRM, and multi-channel sales automation.
- My principle: the project is the purpose; technology is the tool. As technology advances, learning speed and capabilities keep expanding — but what matters is the project to be solved, not the stack.
Experiences
- Lydia Cloud & Jewellery (London, UK · Remote; 2025 — present)
Software / Product Developer — mobile (Flutter), web (React/Next.js), and backend (Node.js) products for a cloud-based jewellery logistics platform; B2B dashboards and AI-assisted tooling. - Oil Trans Petrol ve Enerji (Türkiye; 2024 — 2025)
Software Advisor · Technical Authority — primary software advisor for an energy-logistics operation. Provided management and software solutions for fleet telemetry (ATS) integration, operational CRM for shipment workflows, and invoicing automation. Advised on and coordinated the operational software infrastructure supporting real-time GPS monitoring across a large tanker fleet in the Gabar corridor. - LB Bilişim Teknolojileri (İzmir, Türkiye; 2023 — 2024)
Software Developer — CRM, e-commerce, and admin systems. Led the LB Configurator, an interactive 3D vehicle-and-wheel visualizer with 360° rendering; presented at a German software seminar. - Freelance Full-Stack Developer (Remote · Türkiye; 2021 — 2023)
Delivered 10+ production web applications across real estate, construction, medical tourism, and e-commerce — end-to-end from requirements to deployment.
Education
- Ankara University (Open Education; in progress)
Computer Technologies Web, Associate Degree (2-year) — planning a transition to a 4-year bachelor's program. - IELTS B2 (English)
Selected Projects
Computer Vision & Machine Learning
Sonar-Based Underwater Threat Detection with YOLOv8
- End-to-end deep-learning prototype for detecting divers, ROVs and suspicious underwater objects in Multibeam Forward-Looking Sonar (MFLS) imagery, trained on the open-source UATD dataset. YOLOv8-nano fine-tuned with transfer learning — mAP@0.5: 0.959, Precision: 0.945, Recall: 0.926 across 10 underwater target classes.
Sonar Object Segmentation — YOLOv8x + SAM Hybrid Pipeline
- Hybrid detection + segmentation system for underwater objects in side-scan sonar imagery. YOLOv8x handles detection and classification; Meta SAM produces pixel-precise masks from bounding-box prompts. Trained on SeabedObjects-KLSG — mAP50: 0.801, Precision: 0.936. Includes Streamlit dashboard and batch analysis pipeline.
SAM2-Assisted Rim Segmentation
- SAM2-assisted rim segmentation — foundation models combined with classical computer vision for semi-automatic data generation and custom model training. Achieved 98.7% accuracy with minimal labeled data.
Industrial Wheel Inspection and Feature Recognition
- Real-time computer vision system for a manufacturing facility — automatically detects and classifies bolt holes (lug count) and rim characteristics on the production line without manual intervention.
UrbanSound8K Sound Classification
- Comparative segment-based analysis of urban sounds (car horn, construction, speech) using YAMNet, ESResNeXt, and Audio CLIP — best accuracy 84.1% (ESResNeXt).
AI & LLM Systems
THH-AI — Consumer Arbitration Committee AI Assistant
- AI assistant aligned with Turkish consumer legislation — transforms complaints into legislation-compliant applications, generates unbiased preliminary reports, asks targeted follow-up questions, and references relevant legal provisions.
EmlakRadar — AI-Powered Real Estate Aggregator
- AI-powered real estate aggregator and matching system — intelligent listing discovery, comparison, and personalized recommendations using LLM models.
Full-Stack Web Platforms
LB Configurator — 3D Vehicle & Wheel Visualizer
- Interactive 3D vehicle configurator — users select a car and virtually try wheels in 360°. Scraped vehicle parts aggregated via Next.js staging environment with dynamic wheel-to-vehicle mapping.
Intelligent Customer Management System — AI-Powered Sales Automation
- End-to-end intelligent CRM unifying Instagram, WhatsApp, and Messenger into a single inbox. LLM agents handle 24/7 first contact, match customers to real portfolio listings, drive appointment booking, and feed field outcomes back into the pipeline. Post-sale automation covers birthdays, holidays, re-engagement, and new-project alerts. Role-based team access, listing notes for AI sales arguments, calendar workflow, and conversion analytics — designed and built entirely from scratch.
housing4u.ca — Canadian Real Estate Platform
- Real estate listing platform for Canadian users — search, filter, and manage property listings efficiently.
villakonut.com — Turkish Real Estate Platform
- Real estate platform for Turkish users — search, filter, and manage property listings.
englishcenter.ca — Language Platform
- Language-focused platform supporting individuals planning to move to Canada — resources and essential information.
E-Commerce Platform
- Full e-commerce platform — responsive design, authentication, catalog management, shopping cart, and admin dashboard with real-time analytics.
MediCRM — Medical Tourism CRM
- Full-stack CRM for medical tourism agencies — role-based auth, patient tracking, automated follow-ups, dashboards, and Excel export.
b2bminiui.vercel.app — B2B Dashboard
- B2B management dashboard for jewellery wholesalers — orders, catalogue updates, and client interactions in one interface.
esantiyem.com — Construction Platform
- Modern platform connecting service providers and users in construction projects — planning and progress tracking. Full design and development.
