Hi, I'm
Umair Inayat
Computer Vision Engineer & AI Researcher
Building production AI systems at the edge. Specializing in computer vision, NLP, and large language models with 43+ projects shipped.
About Me
Building AI That Works in the Real World
I'm a Computer Vision Engineer at a startup, where I build production AI systems that solve real-world problems. My work spans the full pipeline — from research and model development to deployment and optimization.
Currently pursuing my BS in Computer Science at the University of Management and Technology with a 3.82 GPA, I've been recognized with the Dean's Award and Rector's Award for academic excellence. I also serve as a Teaching Assistant for Deep Learning and Automata courses.
Beyond my day job, I'm an active open-source contributor building tools like DeepContext (hierarchical memory for AI agents) and AgentTrace (observability for multi-agent systems). I lead the AI/ML team at Google Developer Group on Campus UMT and served as a Beta Microsoft Student Ambassador organizing tech summits across 15 universities.
Journey
Experience & Education
My professional journey, academic background, and leadership roles
2026 — Present
Computer Vision Engineer
Startup
Building production computer vision systems and AI pipelines at a fast-paced startup.
Oct 2025 — Present
Teaching Assistant
University of Management and Technology
TA for Deep Learning and Theory of Automata. Assisting with labs in Python, PyTorch, and TensorFlow. Conducting graded assessments and one-on-one guidance for 100+ students.
Jul 2025 — Sep 2025
AI/ML Intern
Texense
Contributed to production-level AI/ML solutions. Worked on model development, optimization, and deployment in a startup environment.
Mar 2024 — Oct 2025
Teaching Assistant
University of Management and Technology
Assisted in teaching OOP, Data Structures, Database Systems, ML/DL, and Operating Systems. Mentored students in complex programming concepts.
Nov 2023 — Feb 2024
Peer Mentor
University of Management and Technology
Peer tutor for Programming Fundamentals in Fall 2023.
Tech Stack
Technical Skills
Technologies and tools I work with daily
ML & AI
Development
Tools & Infrastructure
Portfolio
Featured Projects
A curated selection from 43+ projects spanning research, products, and open source
DeepContext
Hierarchical memory system for AI agents with knowledge graphs, hybrid retrieval (vector + keyword + graph), and Ebbinghaus forgetting curve decay. Full-stack: Python/FastAPI backend with 188 tests, React dashboard.
XDenseQNet
Quantum-Enhanced DenseNet121 for Monkeypox Skin Lesion Classification. Research paper combining quantum computing with deep learning for medical image analysis.
AgentTrace
Open-source observability platform for multi-agent AI systems. See exactly what your agents are doing — every step, every decision, every cost.
Reelmatic
Full-stack SaaS that generates complete YouTube Shorts from a URL or idea. Produces 9:16 MP4, SEO description, hashtags, and thumbnail.
Specter AI
Open-source, privacy-first AI screen & meeting copilot. Transparent always-on-top overlay invisible to screen share, real-time OCR, audio transcription.
XAI framework achieving 98.4% binary and 86.8% 12-class accuracy using RoBERTa embeddings with stylometric features, LIME, SHAP, and Captum.
QLoRA fine-tuning pipeline for natural text generation. Downloads human-written text from 5 HuggingFace datasets, trains LoRA adapter on Llama 3.2 3B.
GPU kernel programming challenge. Writing CUDA kernels from scratch to build deep understanding of GPU computing.
Full C++ compiler with all 6 phases: lexical analysis, parsing, semantic analysis, IR generation, optimization, and x86-64 assembly output. 4000+ LOC.
AI-powered web app integrating ArXiv, Google Scholar, PubMed, IEEE for comprehensive paper search with LangChain and Groq API.
Deep learning with Vision Mamba and UNETR architecture for retinal vessel segmentation in medical imaging.
PyTorch ViT implementation with Tiny/Small/Base variants, Mixup/CutMix augmentation, LR scheduling on CIFAR-100.
YOLOv11 face detection + EfficientNet V2 classification with MTCNN preprocessing and real-time pipeline.
Multilingual document understanding with translation, NER, sentiment analysis & RAG-based Q&A.
AI-powered social media post generation and scheduling tool.
Training agents to play Flappy Bird using NEAT evolutionary algorithm vs reinforcement learning baseline.
Recognition
Achievements & Certifications
Competition placements, academic awards, and professional certifications
Competitions
TechnoVerse Comsat ML Competition
66 tree species classification with 118k instances
2025
SofTech 2025 ML Challenge
Hospital readmission prediction — FAST Lahore, national-level
2025
Awards
Dean's Award
Fall 2023
Rector's Award
Fall 2024 — 4.0 GPA
Merit Award
Spring 2025
Z.ai for Startups
Accepted 2026
Certifications
Machine Learning Specialization
Andrew Ng / Coursera
Deep Learning Specialization
Coursera
Data Scientist in Python
DataCamp
MLOps Bootcamp
10+ End-to-End ML Projects
Get in Touch
Let's Build Something Together
Open for CV/ML engineering roles, research collaborations, and open-source projects. Feel free to reach out!