Quick Talk
2023A conversational companion built on GPT-4 with retrieval so it answers from real documents. Wired up to Grafana so I could watch it behave in production.
GPT-4RAGMonitoring
I build AI that ships
I'm Reza Amini Gougeh. I've built AI for hospitals, for big tech, and for my own startup, which I sold. RAG pipelines, voice AI, deep learning in production. Based in Montréal, working on what's next.
00 · Start here
Ideas are cheap. Working software isn't.
I'm Reza Amini Gougeh. I take AI ideas and turn them into things people actually use: a voice assistant a patient can talk to, a pipeline that catches Parkinson's in a recording, a search engine that cites its sources instead of making them up.
Most of my work has been in health. Hospitals in Montréal and Ottawa, a research consortium spanning ten institutions, a medical AI company. I've also spent a year at Huawei building smart-home features, and a stretch running my own company until I sold it.
I move between fields on purpose. Health, consumer hardware, and more recently legal tech and large document corpora. Every niche breaks your assumptions in a different way, and you only learn the new thing by being somewhere unfamiliar. The tools carry over. The instincts have to be rebuilt each time.
I care about the boring parts too. Data that's clean, models you can explain, systems that stay up. That's usually where a good demo turns into a real product.
Reza Amini Gougeh
Montréal, QC
- Now
Founder, building something new in legal tech
- Shipped at
- HuaweiJewish General HospitalAmplifier HealthJuztina
- Also
- One startup, built and sold
- 15 papers on Google Scholar, 7 with DOIs listed here
- Speaks
Azerbaijani, Persian, English, French
01 · Work
Where I've built things
Seven roles, mostly in health AI and applied research. The common thread is getting models out of notebooks and into production.
- Early-stage work on an AI product for legal teams. Too soon to say much publicly.
- Same pattern as last time: build the thing, put it in front of people, keep what works.
- Built the AI pipelines and data infrastructure behind the platform, and tuned LLM performance in production.
- The AI features cut churn by 10% and lifted feature adoption by 25%. Deployed on Azure and AWS EC2.
- Shipped a full-stack AI product on my own: web platform, voice AI, audio features, plus RAG pipelines and a citation engine over a large document corpus.
- Made every technical call, then sold the company and the technology behind it.
- Built the data pipelines and AI infrastructure, and put deep learning models into production for medical use, including Parkinson's detection.
- Ran the stack on GCP Cloud Run and AWS EC2.
- Led the AI work for precision medicine in the mental health department at the Lady Davis Institute.
- Took models from research question to something clinicians could actually use.
- Built AI and IoT features for phones, homes, and cars, and cut system latency by 20%.
- Prototyped ideas, demoed them, and contributed to patents.
- Built a virtual companion for patients in Unity, running on the ChatGPT API.
- Improved speech recognition and text-to-speech, and engagement went up 55%.
- Worked on DREAM BIG, a consortium of 10+ institutions studying how genes, prenatal adversity, and early childhood environment shape children's wellbeing.
- Built federated-learning decision support systems and hardened the models to keep sensitive data safe.
02 · Projects
Things I made because I wanted them to exist
Some were research, some were weekend builds, some turned into products. All of them ran.
Keira
2023An AI companion for dementia patients at Bruyère, with ElevenLabs voice and end-to-end encryption. Built for people who need patience more than speed.
Voice AIHealthEncryption
03 · Writing
Notes from the build
Model architectures, things I got wrong, hackathon writeups, and the occasional post about moving to a new country.
05 · Publications
Peer-reviewed work
A systematic review, a multisensory study, an instrumented headset, and the sound-detection work behind ImmerseGuard.
- 01Systematic Review of IoT-Based Solutions for User Tracking: Towards Smarter Lifestyle, Wellness and Health ManagementSensors2024
- 02Optimizing Auditory Immersion Safety on Edge Devices: An On-Device Sound Event Detection SystemOdyssey2024
- 03Multisensory Immersive Experiences: A Pilot Study on Subjective and Instrumental Human Influential Factors AssessmentQoMEX2022
- 04Towards instrumental quality assessment of multisensory immersive experiences using a biosensor-equipped head-mounted displayQuality and User Experience2023
04 · Skills
What I work with
Tools I've actually shipped with, not a list I collected.
- Machine learning
- Python, PyTorch, TensorFlow, scikit-learn, pandas, NumPy, SHAP, LIME
- LLMs and RAG
- RAG design, LangChain, LangGraph, LangSmith, OpenAI, Gemini, Claude, Retell.ai voice AI
- Data and storage
- PostgreSQL, Supabase, Milvus, Qdrant, pgvector
- Shipping and ops
- Docker, Kubernetes, GCP, AWS, Azure, CI, MLflow, DVC, Weights & Biases
- Product surface
- FastAPI, Flask, Django, Streamlit, Stripe, Unity
07 · Contact
Say hello
If you're building something in AI and want another pair of hands on it, or you just want to talk about a problem, reach out on LinkedIn or GitHub. I answer.