Aziz Ahmed
Software engineer working on machine learning, healthcare, and the systems that put models into production — with a growing focus on AI security.
SupportX
2025An Azure RAG platform that turns scattered docs into searchable context. Reduced L1 ticket volume ~60% and brought response time from ~25s to ~1.3s.
Churn Prediction Pipeline
2025A self-retraining churn system on Azure MLOps that only spins up GPUs when needed — experiment tracking, reproducibility, and Deep Kernel Learning included.
BurnOut
2025A multi-agent workout planner that generates plans, macro targets, and progressive adaptation from check-ins — modular agent routing with persistent profiles.
Offensive Security Assessment
A full-scope web application penetration test covering OWASP Top 10 — authentication bypass via SQL injection, manual and automated SQLi, stored/reflected XSS, local file inclusion, web-shell upload, and unvalidated open redirects. Ran OSINT recon to map the attack surface, chained findings into realistic impact rather than isolated PoCs, and delivered a 50+ page report with annotated evidence, severity ratings, and prioritized remediation.
VIAstra — A consultant-led AI product for smarter cervical screening
SupportX — Engineering a retrieval-augmented AI system
MLOps in the real world — GPU-driven Deep Kernel Learning for churn
A clinical NLP pipeline that touches an LLM exactly once
Lead Developer
Turning my BRAC research into a two-stage clinical support system: image screening plus risk prediction. Fine-tuned ResNet-50 with transfer learning to classify VIA screening images as positive or negative, and developed an ensemble model on patient clinical data that flags VIA-negative patients at risk of turning positive. Built the end-to-end pipeline in Python with PyTorch, TensorFlow, and Keras, from ingestion to inference. Tested with a partner clinic during research; now improving accuracy and robustness for clinical use.
Software Engineer Intern
Built software features in C++ from business requirements and technical specifications, working across the full lifecycle: requirements analysis, design, development, testing, deployment, and maintenance. Optimized application components for speed and responsiveness across devices and browsers, set up GitHub Actions CI/CD with automated tests and review gates, and containerized services with Docker. Worked with senior developers to trace pipeline failures to their root cause.
Artificial Intelligence Research Assistant
Researched deep learning for cervical cancer screening from VIA images, the work that later became VIAstra. Curated raw clinical image datasets into standardized, model-ready formats and built preprocessing and feature-extraction pipelines that raised model F1-score by 14%. Ran ablation studies to test which data and model choices mattered, and documented the pipeline for an internal publication and the lab’s open-source prototype.
Cybersecurity Intern (VAPT)
Performed penetration tests on 3 web applications, from OSINT reconnaissance to exploitation. Exploited OWASP Top 10 flaws including SQL injection login bypass, XSS, local file inclusion, and web shell upload, using Kali Linux, Burp Suite, SQLMap, OWASP ZAP, Acunetix, and Nmap for scanning and API fuzzing. Wrote a 50+ page assessment report with step-by-step attack evidence, severity ratings, and prioritized fixes.
M.S., Cybersecurity — AI Security
B.Sc., Computer Science & Engineering
OCI 2025 Certified Generative AI Professional
Oracle · Oct 2025
Outside the day job, I co-founded Vision360, a non-profit educating street children in Bangladesh, and I mentor developers 1:1 on Fiverr and Upwork. When I step away from the keyboard, you’ll usually find me playing guitar.