I founded an AI startup, built the whole thing myself, and sold it along with its technology. Web platform, voice AI, RAG pipelines, citation engine, cloud infra. All of it.
That is the short version. The longer one: six years of applied AI across healthcare, human computer interaction, and LLM products, with a full stack background that means I own the model, the pipeline, the API, the infra, and the interface a real person touches.
I like problems where the signal is messy and the user is real. Voice as a vital sign. EEG as a game controller. A chatbot that has to be kind to a dementia patient.
| 🚀 | Founded and sold an AI startup. Full stack product with voice and audio AI, RAG over a large document corpus, citation engine. Every technical call was mine. |
| 🧠 | Interim AI Lead at Juztina. Production LLM pipelines on Azure and AWS. Churn down 10%, feature adoption up 25%. |
| 🎙️ | AI Engineer at Amplifier Health. Deep learning for disease detection from voice, deployed on GCP Cloud Run. |
| 🏥 | Lead ML Engineer at CIUSSS West-Central Montreal. Precision medicine AI, federated learning across 10+ institutions, 30% faster deployments. |
| 📍 | Montreal, QC. Open to conversations about applied AI, LLM infrastructure, and health tech. |
Languages
AI, LLMs and machine learning
Frameworks and product
Data, databases and vector stores
Cloud and DevOps
MLOps and observability
Version control and collaboration
And the things that do not have a logo
| Area | What I use |
|---|---|
| LLM engineering | RAG pipeline design, citation engines, domain tuned workflows, LangSmith tracing, prompt and eval harnesses |
| Voice and audio | Retell.ai, ElevenLabs TTS, ASR and transcription, meeting intelligence, real time audio pipelines |
| ML tooling | PyCaret, NLTK, SHAP, LIME, Evidently AI, federated learning, model monitoring |
| Data and viz | Matplotlib, Seaborn, SPSS, biosignal and medical image processing |
| Product | Supabase Auth with OAuth and SSO, Stripe billing and subscriptions, Vertex AI, SageMaker, Arduino |
| Ways of working | MLOps and LLMOps practice, cross institutional research collaboration |
| Role | Where | When |
|---|---|---|
| Founder & Applied AI Engineer | Stealth AI Startup, acquired | 2025 – 2026 |
| Interim AI Lead | Juztina LLC, remote | 2025 – 2026 |
| AI Engineer | Amplifier Health, remote | 2025 |
| Lead Machine Learning Engineer | CIUSSS West-Central Montreal, Lady Davis Institute | 2024 – 2025 |
| HCI Research Engineer | Huawei Technologies Canada, Markham | 2023 – 2024 |
| Full-Stack Dev & AI Specialist | UQO, Elisabeth-Bruyere Hospital, Ottawa | 2023 |
| Machine Learning Engineer | CIUSSS West-Central Montreal, Jewish General | 2021 – 2023 |
What that actually looked like
Stealth startup. Architected and shipped a full stack AI product: web platform, voice AI, audio features. Built RAG pipelines, a citation engine, and domain tuned LLM workflows over a large document corpus. Led every technical decision across AI architecture, cloud, and reliability. Sold the company and its technology.
Juztina. Designed and maintained scalable AI pipelines and data infrastructure, optimizing LLM performance in production. Shipped AI features that cut churn 10% and raised feature adoption 25%. Ran services across Azure and AWS EC2.
Amplifier Health. Put deep learning models into production for real medical use, including Parkinson's detection from voice. Deployed on GCP Cloud Run and AWS EC2.
CIUSSS and Lady Davis Institute. Led precision medicine AI in the mental health department. Built automated ML pipelines that improved deployment efficiency 30%. Developed federated learning over multimodal clinical data with 10+ global institutions in the DREAM BIG consortium. Hardened models against data leakage.
Huawei. Applied AI for IoT and wearables in smart homes. New features for phones, homes, and cars with 20% lower latency. Prototyped and demoed concepts for product integration and contributed to patents. Won the Outstanding Team Award of the Canadian Research Centre.
UQO and Bruyere. Built a virtual companion for patients in Unity on the GPT-4 API. Improved speech recognition and text to speech, raising user engagement 55%.
|
Conversational AI you actually want to talk to. GPT-4 with dynamic prompt engineering, RAG for context, Grafana for live health and performance monitoring. |
A ChatGPT powered companion for dementia patients at Bruyere Hospital. ElevenLabs TTS for a voice that sounds like it cares, end to end encryption because the data is sacred.
|
|
A multiplayer game for kids with disabilities, driven by a real time SSVEP brain computer interface. Built at BCI Game Jam. 🥇 1st place, Popular Vote, Best SSVEP Game |
Master's thesis. Multisensory VR training to improve motor imagery BCI accuracy, where machine learning, biosignal processing, and UX meet.
|
| 🎓 M.Sc. Telecommunications | INRS, Universite du Quebec. Thesis on multisensory VR training for motor imagery BCI. |
| 🎓 B.Sc. Biomedical Engineering | Medical image enhancement, multimodal biomedical data, signal processing. |
| 🏆 Awards | Huawei Outstanding Team Award, McGill MEITA, McGill Graduate Excellence Fellowship, UQ International Tuition Exemption, BCI Game Jam 🥇 |
| 📜 Certificates | MLOps Concepts, Google Project Management, AWS Generative AI with Diffusion Models, Developing Generative AI Solutions |
| 🌍 Languages | Azerbaijani (native), Persian (native), English (fluent), French (beginner) |
Building something interesting? Let's talk.
magnumical.ca · LinkedIn · X