Citizens of Silopolis: Pedram Ghazi, AI Engineer

Pedram Ghazi is an AI Engineer at Silo AI, specialized in computer vision. So far, Pedram has worked in computer vision projects to build object detection and visual tracking solutions. He has also been a part of our internal R&D projects, including internal infrastructure and software library work.

Prior to joining Silo AI, Pedram finalized his Master of Science degree in “Data Engineering and Machine Learning” at Tampere University. After graduation, he continued his research in the robotics area, focusing on the environment perception for heavy vehicles before moving to Helsinki to join team Silo.

Giving “sight” to machines with computer vision

At Silo AI, we build AI solutions that help us humans become better at their work. One of the key AI technologies in such solutions is computer vision, which makes it possible to draw machine-driven insights and data based on images and videos. Pedram is excited for his work as a computer vision expert for a good reason:
“Put to simple terms, we are making computers able to recognize and to remember different things and objects! How cool is this to you?”

The process of giving “sight” to machines is all about developing different algorithms and extracting data from various sensors.

“As a computer vision expert, I’m trying to develop methods and algorithms to help computers 'see' and comprehend the surrounding world like we people do. Once the understanding gets more accurate, people can start to work with the assumptions made by the computer. An example would be a port operator that works with an AI solution that gives an up-to-date situation of all the cargo ships in the port."

Custom-made solutions that combine research and domain expertise

Improving situational awareness is a typical goal for a computer vision solution, be it about identifying defects in a production line or classifying cancer cells from a biopsy image. "Many computer vision solutions tackle problems that have existed all the time, but so far there hasn't been an effective solution." Pedram explains.

As the computer vision solutions are typically built for a specified environment, making the AI solution recognize the correct things requires a significant amount of domain expertise:

“As an AI Engineer I seek to truly listen to the client and think about their working environment. We have to understand their domain as we train the model. However, understanding the organizational workflows too is crucial to build an AI solution that is adopted well. That is why one of the critical roles in my job is to create and maintain a collaborative connection with customers.” Pedram continues.

The solutions Silo AI builds are often research-based AI models and therefore involve a lot of testing and development before being implemented into the actual pipeline. Out of Silo AI’s growing team of 80 employees, nearly half have a PhD degree in machine learning or other relevant field. Pedram enjoys this dynamic of constant research and application in his work:

“As we provide novel and tailored AI solutions to businesses, my job as an AI Engineer requires constant learning. Our part is to stay on top of the technology and research, and combine that with our client's domain expertise.”

Trust is built through expertise and communication

Silo AI aims at being a trusted advisor in all things AI for its clients. Depending on the client's maturity in terms of data and digitalization, an AI project can involve a high amount of data pre-processing and collecting. To do this properly, Pedram highlights the importance of trust:

“In many cases, our customers are not necessarily fully aware of what is actually causing the issue we’re trying to solve. Sometimes, they’re looking at the right problem but are unsure of how to deal with the data they have. When trouble-shooting data related issues, it is important to be trusted with all the relevant information."

Communication is a key element in building and maintaining trust: clarifying the problem for all the stakeholders at the early stages of the project can help the future of the project. An AI solution rarely impacts just a single part of the organization, but it rather has potential to improve entire workflows.

“Being agile and making your research properly helps in making smarter choices. You need to understand the problem, look for similar issues and find suitable solutions, implement them, deliver and deploy them. That’s an AI project in short,” Pedram continues.

Focus on the field of interest in projects

At Silo AI, most of the AI experts are able to leverage their core fields of interest and expertise in the projects. With initiatives like talent coaching and research club, Silo AI supports the team's learning.

“I’m continuously investigating new ways to solve these challenges, when using modern computer vision techniques. The field advances very fast and it’s important to keep track of the recent papers. 

“At Silo AI, I've been able to focus on real-life use cases where I can apply my knowledge to a concrete case. It's been rewarding to stay more focused and improve in my own line of work. ”

To put more technically, Pedram has been able to apply some of the key technologies in computer vision that he was studying and researching during his studies at Tampere university: "Some of the methods I’ve developed for the object detection or tracker pipelines make use of 2D RGB monocular cameras, stereo images, and 3D point clouds," Pedram comments.

Computer vision was a natural choice 

Pedram has always been wondering about how computers and machines could perceive their own surroundings and environment. This originally made him interested in computer vision, and to follow his passion:

“Look at me now! I’ve been working with exciting cases where I’ve got to build algorithms that identify and re-identify vehicles or people in different scenes. I really enjoy working in the field of computer vision since it’s put me on the path to solve the problems that have intrigued me for long. Analyzing the data coming from different sensors like cameras, LiDAR and radars is a way to enable computers to 'see' and to enable people to 'see more effectively' through these smart solutions”

Pedram appreciates also the help he gets from his colleagues:
“I enjoy the friendly and warm environment we have at Silo AI. I couldn't be more grateful for the technical but also mental support we share internally. Everyone is acknowledged and appreciated.”

“I like to bring joy and cheerful energy to the group since the energy gets amplified and comes back to myself. At Silo AI, we work in different teams in close collaboration. This sense of integrity between all the team members helps a lot in having a positive and productive environment."

Free time

During his free time Pedram likes to socialize and hang out with his friends. As for sports, he likes to go to the gym. Pedram enjoys relaxing by watching his favorite TV series, and listening to music, audiobooks and podcasts.

Favorite Silo AI value

When asked what's his favorite Silo AI value out of Keep Learning, Be Good, Build Bonds and Ask Why, Pedram says:
“All of Silo’s values look very promising to me, they are deeply intertwined with each other. But the boldest one to me is the 'Keep Learning'! Silo knows that an employee’s growth is equal to its own growth and is constantly helps and guides their employees towards improving themselves in different aspects.”

Pedram’s tips:

If you're into computer vision and machine learning, Pedram collected a nice set of tips that have helped him in his work as AI Engineer.

• Blog posts in e.g. Medium and Towards data science websites which are covering and explaining many complicated topics and make them easier to understand.
• Finding related topics on the paperswithcode.com website which is very well suited for both finding new approaches and for making comparisons between these methods.
• Kaggle is a well-known website that holds machine learning competitions. I use it for collecting information and ideas about different areas in computer vision, but also to access open datasets. It's also fun to talk with my colleagues about some of Kaggle competitions they've joined.
• Different scientific channels on social media like LinkedIn, Facebook, Telegram. You can find many channels in them which are posting the latest inventions and discoveries in different fields.

Would you like to work with AI Engineers like Pedram? Check out our open positions at https://silo.ai/careers.