Hey, I'm Mehshid Atiq
Engineering little sparks of intelligence from pixels, data, and stubborn code...
About Me
I am an AI/ML Engineer who combines full-stack software development with advanced machine learning. Leveraging expertise across LLMs, multimodal AI, computer vision, and human-centric systems, I build scalable, context-aware applications. At my core, I design intelligent software engineered to minimize loss and maximize impact.
Education
Msc. AI and ML
Technical University of Darmstadt
Oct 2025 - Present
B.S. Computer Engineering
Bilkent University
Aug. 2021 - Jun. 2025
Experience
Software Developer Intern
Smarnovative Labs
Jul. – Sep. 2024
Backend Developer Intern
Creative Garage
Jul. – Sep. 2023
Professional Experience
Research Assistant
- Built a multimodal emotion-recognition pipeline for 26-class EMOTIC classification using facial and scene context.
- Benchmarked ResNet-50, ViT, and CNN-based models to compare architecture performance across visual emotion tasks.
- Improved F1/mAP under class imbalance through loss weighting, threshold tuning, and ablation-driven optimization.
Software Developer Intern
- Developed real-time performance dashboards with Python, BigQuery, and Tableau—40% issue identification time reduction.
- Implemented data ingestion pipelines and automated integration using Python, Cloud Functions, and Pub/Sub.
- Conducted comparative analysis of Unity vs. Unreal Engine for improved rendering and tool integration.
Backend Developer Intern
- Enhanced ERP system modules for property management and automated billing using Yii2 and MySQL.
- Designed secure role-based access portals leveraging Yii2’s RBAC for optimized data control.
- Performed rigorous testing and code reviews to ensure robust, high-performance backend solutions.
Featured Projects
OpsPilot
AI operations copilot for German IT and consulting SMEs that automates admin-heavy work through modular skills. It uses company knowledge to process emails, meetings, and documents, then proposes follow-ups, draft replies, tasks, and next steps while humans approve every sensitive action.
Emotion Recognition System
Multi-model system classifying 26 emotions from the EMOTIC dataset using ResNet-50, Vision Transformers, and a custom CNN. Features adaptive thresholding and dynamic weight adjustments.
Smart Vector Query (SVQ.ai)
Developing an AI tool using Retrieval-Augmented Generation (RAG) with vector databases and semantic search for precise info retrieval.
TD3 Reinforcement Learning Agent
TD3 agent for LunarLanderContinuous-v3 with custom actor-critic networks, dual critics, and Gaussian noise-based exploration for continuous policy optimization.
Campus Connect
Web application for university marketplace transactions and community engagement. Features Elasticsearch, JWT security, and a scalable Docker-based backend with Minio.
Notes Muscle
Android app for lecture recording and note sharing focused on privacy. Built with a custom TCP protocol and multithreaded server for optimized binary file management.
Core Expertise
Professional proficiency across the AI/ML and software engineering landscape.
AI & Machine Learning
Data Engineering
Full-Stack Web
Data & Vector
Cloud & DevOps
Mobile & Systems
Recent Insights
The Future of Generative AI in Software Development
Exploring how LLMs are reshaping the coding landscape, from autocomplete to autonomous agents.
Read ArticleBuilding Scalable Microservices with Next.js
A deep dive into server actions, edge functions, and how to architect robust applications.
Read ArticleUnderstanding Reinforcement Learning
Breaking down the basics of RL agents, rewards, and environments with practical examples.
Read Article