Syed Burhan Ahmed
Building production-grade ML systems using Python, PyTorch, YOLOv11, FastAPI, LangChain, and n8n. Currently shipping a production recommendation engine and an AI vendor-search chatbot at OptimusFox, architecting a multi-tenant B2B SaaS platform, and benchmarking VLMs against human baselines for my Final Year Capstone Project. IEEE Published Author (ICIC 2025).
About Me
AI Engineer & Researcher at UMT Makerspace Lab — building production-grade CV and GenAI systems that bridge research and deployment.
My Journey
I am an AI Engineer building robust, production-grade systems across Computer Vision, Agentic AI, and applied ML. At OptimusFox, I built and shipped BookItUp's recommendation system — running sub-300ms, frequently sub-100ms, in production. I'm now building an AI-powered vendor-search chatbot, architecting a multi-tenant B2B SaaS product, and designing an n8n-based marketing automation pipeline, while leading a team of 6 AI interns and coordinating directly with the Engineering Manager on the AI team.
In research at UMT's AVR Makerspace Lab, my FYP benchmarks closed-source vs. open-source VLMs — Gemma 4, Grounding DINO, CLIP, and Jina — against human baselines on real-world perception tasks including object counting, license plate recognition, and scene understanding, all on a constrained 16GB GPU. I also developed InsightVision, a FastAPI microservice orchestrating a hybrid detection-and-tracking pipeline (Grounding DINO + BoT-SORT, SAM3 + BoT-SORT, and standalone SAM3 configurations, with Gemma 4 driving OCR and prompt-driven querying) achieving <100ms latency for semantic natural-language visual search.
I published my first IEEE paper at ICIC 2025 on Vision-Based Fall Detection for Elderly Surveillance using YOLOv11 (DOI: 10.1109/ICIC68258.2025). Beyond engineering, I have trained 120+ participants as a Master Trainer in Generative AI and served as Teaching Assistant across three semesters covering Digital Logic Design and Introduction to Robotics.
Open to collaborating on challenging, impact-driven engineering problems. Whether it’s deployable computer vision systems, workflow automation infrastructures, or mentoring developers, the focus is on building reliable tech that solves real-world bottlenecks.
How I Work
Clean Code & Practical AI
Writing scalable, production-ready AI systems with an emphasis on maintainability and real-world impact.
Continuous Learning
Staying ahead of research trends, exploring new architectures, and pushing boundaries in CV and GenAI.
Collaboration & Mentorship
Working with teams, mentoring engineers, and building clear documentation to multiply collective expertise.
Performance & Automation
Optimizing inference pipelines, automating workflows, and delivering high-performance AI systems in production.
Technical Expertise
Battle-tested in production ML systems, real-time vision pipelines, and GenAI applications.
Languages
7 skills
Python
C/C++
C#
JavaScript
MIPS Assembly
SQL
HTML/CSSProfessional Journey
From enterprise solutions to tech innovation, building scalable systems and leading AI-driven initiatives across diverse technology stacks
AI Engineer
OptimusFox
Building production AI systems across recommendation engines, agentic chatbots, and workflow automation for live products.
Key Achievements
- Built and shipped BookItUp's production recommendation system, consistently running sub-300ms and frequently sub-100ms at scale.
- Building an AI-powered vendor-search chatbot for BookItUp, enabling natural-language discovery and tool-calling over existing REST APIs.
- Designing an n8n-based marketing automation pipeline with a custom UI layer for a scalable B2B SaaS product.
- Architecting a multi-tenant B2B SaaS platform while leading a team of 6 AI interns, coordinating directly with the Engineering Manager on the AI team.
Technologies Used
AI Researcher – Computer Vision & Robotics
AVR Research Group Makerspace Lab, UMT
Conducting applied research in VLM benchmarking, hybrid vision pipelines, and multimodal perception for real-world deployable systems.
Key Achievements
- Benchmarking closed-source vs. open-source VLMs (Gemma 4, Grounding DINO, CLIP, Jina) against human baselines on perception tasks — object counting, license plate recognition, scene understanding — on a constrained 16GB GPU.
- Developing InsightVision, a FastAPI microservice orchestrating a hybrid detection-and-tracking pipeline (Grounding DINO + BoT-SORT, SAM3 + BoT-SORT, standalone SAM3) with Gemma 4 handling OCR and prompt-driven querying, achieving <100ms latency for semantic natural-language visual search.
- Finalized the InsightVision backend architecture and system documentation ahead of final FYP defense, with frontend deployed on Vercel and backend deployment via Tailscale Funnel/Cloudflare Tunnel underway.
- Published at IEEE ICIC 2025 (Dec 2025) on Vision-Based Fall Detection for Elderly Surveillance using YOLOv11. DOI: 10.1109/ICIC68258.2025
Technologies Used
Teaching Assistant – Digital Logic Design & Introduction to Robotics
University of Management and Technology
Assisting course instructors in delivering Digital Logic Design and Introduction to Robotics by managing course operations, assessments, and student support across multiple semesters.
Key Achievements
- Serving as TA across 3 semesters: Digital Logic Design (Spring '25, Fall '25) and Introduction to Robotics (Spring '26).
- Managed LMS for 100+ students per semester including assignments, quizzes, and project submissions.
- Maintained gradebooks, coordinated student queries, and ensured timely course delivery.
Technologies Used
Data Systems & Automation Associate
Punjab Skills Development Authority (PSDA)
Optimizing backend data integrity and system performance for the Single TVET-MIS, ensuring scalability for statewide vocational education programs.
Key Achievements
- Managed data workflows for a system serving 100,000+ concurrent users, ensuring high availability and query optimization.
- Automated data reporting pipelines for 1,100+ institutes, reducing manual data entry overhead and ensuring 99.9% data accuracy.
- Collaborated with engineering teams to maintain backend stability and implement role-based access protocols using SQL Server.
- Executed rigorous data cleaning and transformation scripts to standardize provincial vocational education datasets.
Technologies Used
Software Development Intern
BeWhoop
Selected as a Software Development Intern (~20–25 hours/week) focusing on .NET technologies and contributing to core software projects in a fully remote environment.
Key Achievements
- Gained hands-on experience with .NET development, building and integrating new software modules within existing systems.
- Demonstrated strong problem-solving skills and contributed creative solutions while collaborating virtually with the engineering team.
- Ensured timely completion of assigned tasks, consistently meeting project quality standards and deadlines.
- Maintained professional communication across distributed teams and time zones, supporting smooth remote collaboration.
- Actively participated in remote team discussions, sharing insights to enhance productivity and code quality.
Technologies Used
Master Trainer – Generative AI
Arfa Karim Technology Incubator (AKTI)
Designed and delivered workshops on Generative AI, large language models, and prompt engineering for early-career tech professionals and students.
Key Achievements
- Developed a practical curriculum and interactive sessions for 120+ participants, covering modern LLM and generative AI workflows.
- Mentored cohorts on developing innovative project submissions that applied generative AI to real-world use cases.
Technologies Used
Freelance Software Engineer
Lahore Arts Council
Led development of a reservation web system for event space management, including admin dashboards, user portals, and automated email-based workflows.
Key Achievements
- Deployed a .NET/SQL Server-based web reservation system with integrated Excel reporting to support operational decision-making.
- Achieved a largely automated reservation workflow serving 500+ monthly users.
Technologies Used
Featured Projects
AI-driven systems for vision, language, and automation—integrated for real-world education, industry, and innovation.
InsightVision
Offline AI Video Surveillance & Semantic Search Platform
A fully offline AI video surveillance platform built as my Final Year Capstone Project, orchestrating a hybrid detection-and-tracking pipeline (Grounding DINO + BoT-SORT, SAM3 + BoT-SORT, and standalone SAM3) with Gemma 4 handling OCR and prompt-driven querying, achieving <100ms latency for semantic natural-language visual search. Delivered as a FastAPI microservice with a React 'Cyber-Ops' dashboard and Redis-backed state, developed under UMT's AVR Research Group.
BookItUp Recommendation Engine
Production AI Recommendation System
Designed and shipped a production recommendation system for BookItUp, a live booking platform, consistently returning results sub-300ms and frequently sub-100ms at scale. Integrated end-to-end with the existing backend and validated against production traffic.
RULEN
Real-Time Unified Lecture Extraction Network
AI-powered multilingual educational assistant automating lecture transcription (99+ languages), bilingual notes in Roman Urdu/Hindi & English, and curriculum-aligned quiz creation with RAG. Supports dashboards, CLO/PLO mapping, multi-format upload, and integrates with learning outcomes for institution readiness.
Freak AI
Memory-Augmented Conversational Agent
A rule-based conversational agent modeling five human-inspired memory systems (social, semantic, PAM, episodic, sensory) as a live Neo4j knowledge graph, with real-time D3.js dashboards visualizing memory formation. Combines AIML pattern matching with Prolog fact reasoning, plus an optional Alexa-style ESP32 voice interface with DHT11 sensor logging for spoken interaction.
CelebVision – Dual-Shot Celebrity Classification
Multi-Pipeline Face/Gender/Celebrity Recognition
CV pipeline using YOLOv11 for face detection, EfficientNetV2-S for celebrity recognition (15+ Pakistani celebrities). 1st place in CV Kaggle competition; real-time inference, visualizes bounding boxes and recognizes gender, class imbalance handled.
ArUcoNav: Autonomous Vision-Guided Robot
Real-time pose estimation & dynamic path planning
Co-engineered an autonomous navigation stack (Python + OpenCV) using ArUco marker pose estimation, after a rival color-tracking approach failed under venue lighting on demo day. Diagnosed and fixed an inverted heading-vector bug through manual vector-math debugging, then implemented dynamic 'Safety Node' routing via Euclidean distance to steer around obstacles. Deployed on ESP32 with PID steering correction for a clean 13-second finish.
Real-Time Fall Detection System using YOLO
Elderly Safety & Surveillance
YOLOv11-based fall detection system for real-time video feeds in safety/public spaces. Fine-tuned on LE2I, high precision/recall, supports logging and alerts, modular for sensor integration.
Lahore Arts Council Reservation System
Event Space Reservation Web App
Designed and developed a full-stack .NET and SQL Server application for event reservations. Features include user registration, an admin dashboard, automated email notifications, and Excel report generation. Built as a proof-of-concept to demonstrate robust reservation workflows in a simulated environment.
Voice-Based Smart Attendance System
IoT-Driven Attendance Platform
Built the ESP32 control hub for a dual-mode attendance system: passive NFC + fingerprint verification for daily attendance, and a voice-driven admin mode for student enrollment. Captured audio via a MAX9814 mic, sent it to a Flask speech-to-text server over HTTP, and routed the returned keyword (enroll/update/delete/scan) to trigger the matching action — with DFPlayer Mini audio prompts and OLED status feedback throughout.
Interested in collaborating on innovative AI/ML projects?
Publications & Writing
Research contributions and technical deep-dives into CV, NLP, and Optimization algorithms.
Efficient Vision Based Fall Detection for Elderly Surveillance Using YOLOv11
Real-time fall detection system leveraging YOLOv11 and computer vision for elderly surveillance in video-based safety monitoring.
Understanding Principal Component Analysis (PCA) in Machine Learning: A Comprehensive Guide
Explains PCA for dimensionality reduction, covering its mechanics, importance in ML workflows, and practical implementation.
When to Use Mean, Median, and Mode for Handling Missing Values in Data?
A practical guide to selecting statistical imputation methods (mean, median, mode) for handling missing data effectively.
1D Convolutional Neural Networks (1D-CNN): A Powerful Tool for Sequential Data
Explores 1D CNNs for sequential data, covering working principles, advantages, and a hands-on TensorFlow/Keras implementation.
Understanding Gradient Descent in Machine Learning
Explores the mechanics and variations of gradient descent, the fundamental optimization algorithm driving modern machine learning and deep learning.
Understanding Long Short-Term Memory (LSTM) Networks in Deep Learning
Comprehensive guide to LSTMs for NLP and forecasting, detailing their architecture and advantages over basic RNNs.
Understanding Gated Recurrent Units (GRU) in Deep Learning
Covers GRUs as an efficient LSTM alternative, explaining their architecture, mechanics, and deep learning applications.
Professional Certifications
Continuous learning and professional development in cutting-edge technologies
Google Cloud Fundamentals: Core Infrastructure
Coursera
Covers core Google Cloud concepts and services, including organizing infrastructure, creating basic cloud resources, and selecting appropriate storage options.
Committed to staying current with the latest technologies and best practices
Awards & Achievements
Honoring accomplishments in AI research, academic excellence, and impactful technical leadership.
HEC NSCT – Top 0.8% Nationally
Higher Education Commission (HEC) Pakistan
Outperformed 99.2% of 33,038 students across Pakistan in the HEC National Skill Competency Test — AI & Computing track.
Rector's Merit Award – Spring 2025
University of Management and Technology (UMT)
Conferred by the Rector for achieving a perfect 4.00/4.00 SGPA in Spring 2025, in recognition of exceptional academic excellence.
Dean's Merit Award – Spring 2023
School of Systems & Technology, UMT
Conferred by the Dean for earning a high semester SGPA and ranking among the top-performing students in the Artificial Intelligence program.
Gold Medalist – Matriculation
Lahore Garrison Education System
Secured the Gold Medal at the Matriculation level for outstanding academic performance.
Get In Touch
Ready to bring your ideas to life? Let's discuss your next project