Haim Cohen
Hi, I’m Haim—a Data Scientist and Machine Learning Engineer with over five years of experience in developing AI tools that make a tangible impact.
I’ve had the chance to work across gaming, business communication, and fintech, where I’ve focused on applying my expertise in machine learning and data analysis to create systems that are both practical and reliable. I love diving into complex challenges, no matter the field. Whether it's exploring new technologies or optimizing existing systems, I’m all about hands-on work, collaboration, and continuous learning.
My goal is simple: to build solutions that work and make a real difference. If there’s a challenge ahead, I’m ready to tackle it head-on.
Career Journey
Computer Vision Engineer at ShapeShift Gaming
At ShapeShift Gaming, our mission was to revolutionize the live-streaming experience by extracting game elements and UI from real-time creator content to build a recommendation system tailored for platforms like Twitch.
My contributions included implementing OCR with classical CV techniques—such as bitwise operations, contours, and template matching—alongside neural networks (NNs), KNN, and Tesseract, to accurately extract and analyze in-game UI elements. To further enhance the robustness of our models, we created augmented data by simulating game elements in challenging environments (e.g., low bitrate, blurry backgrounds, game effects) and automatically tagged this data, significantly improving our system's accuracy in real-world conditions.
We further enriched our data pipeline by integrating various methods from CV, DSP, and NLP to gather essential information from streamers. For example, in our gender recognition efforts, we used a combination of CV (VGG-Face DCNN), DSP (pyAudioAnalysis), and NLP (TF-IDF) to predict and enhance gender accuracy, contributing to more personalized recommendations.
In a specific project involving Fortnite, we needed to determine the number of players on screen to identify the current game mode. To achieve this, we employed a YOLO human detection model, which I finetuned through transfer learning on game character data. This model, along with other optimizations, significantly improved our system’s ability to accurately detect and classify game modes. These efforts led to a substantial improvement in recommendation accuracy, content discovery, and overall viewer engagement.
Deep Learning Researcher at Substrata
At Substrata, we focused on enhancing business communication by analyzing real-time human dynamics to uncover attitudes, intentions, and power dynamics, helping deal-makers sell smarter and close more deals. My primary contribution was the development of a non-verbal LLM that interprets and predicts power dynamics in business interactions.
By analyzing both manually tagged data and real-time interactions, we achieved a 20% improvement in communication effectiveness, directly impacting deal outcomes. In addition to my technical contributions, I played a key role in fostering a culture of meticulous documentation and thorough research. This focus on detailed documentation ensured that our models were not only effective but also well-understood and maintainable, laying the groundwork for long-term success and scalability.
I also developed a regression model to optimize email sending times by analyzing historical interaction data, which resulted in a 15% increase in deal-closing rates. Additionally, I significantly improved the accuracy of our email analysis system by creating a model for email signature detection using high-quality, manually tagged data, boosting the power dynamics score baseline by 40%.
Moreover, I took the initiative to showcase both new and established technologies to my coworkers and stakeholders. This not only allowed me to deepen my understanding of various subjects but also contributed to a culture of continuous learning and innovation within the team.
Throughout my time at Substrata, I was instrumental in integrating AI-driven insights into our platform, enabling users to navigate complex negotiations with greater confidence and precision.
Lead Data Scientist at ThetaMind
At ThetaMind, I co-founded and developed a fintech platform focused on optimizing and scaling a proprietary options trading algorithm. This experience not only deepened my technical expertise but also provided valuable insights into the intricacies of the fintech industry, including regulatory considerations and market dynamics.
increased algorithm performance by 25% through detailed research and identification of key optimization areas using classic statistical analysis and machine learning techniques like regression and random forests, measured using historical market data to enhance early development phases.
As a co-founder, I took on responsibilities beyond the technical realm, from strategic planning to investor relations, gaining a holistic understanding of what it takes to build and scale a startup. I also architected an efficient backend service solution to handle real-time symbol and options data, capable of withstanding 13k requests per minute. This was achieved by leveraging NestJS to build a scalable microservices architecture and deploying on Vercel for optimized serverless performance. The system's reliability was validated through rigorous stress tests under simulated market conditions, ensuring it could perform reliably during live trading sessions.
This experience solidified my passion for combining technical innovation with strategic business insight. It has equipped me with a unique blend of skills that I am eager to apply to new challenges, whether in fintech or beyond. The journey at ThetaMind has been a testament to the power of interdisciplinary collaboration and the impact of well-engineered solutions on real-world problems.
Tech Arsenal
Programming Languages
Frameworks
Databases
Libraries
Models
Infrastructure
Endorsements
Haim is one of those rare talents who brings both deep technical expertise and a genuine passion for innovation to the table. From day one at Substrata, he was more than just a brilliant engineer—he was a driving force that energized our team. His ability to turn complex ideas into practical, impactful solutions is something that still impresses me. Beyond his technical skills, Haim has a knack for fostering a culture of curiosity and continuous learning. He doesn’t just solve problems; he inspires those around him to reach higher and think bigger. Working with Haim was not just productive—it was genuinely fun, and I’m excited to see where his journey takes him next. Wherever Haim goes, he’s sure to make a big impact, and I’m eager to see what he’ll accomplish in the future.
Baruchi Har-Lev
Co-Founder & CTO, SubStrata
It’s hard to do Haim justice with such few characters, but the first words that come to mind are selfless, dedicated, and extremely trustworthy. I hired Haim as an entry-level python engineer and grew to be one of the most phenomenal engineers on our team. His drive and hustle, paired with his infectious passion for gaming and code, made for a winning combo, and he shows no signs of slowing down. It’s been a true honor to work side by side with him, I highly recommend him!
Kevin Edry
Co-Founder & CTO, Shapeshift Gaming