Rofy
AI Ethics and Equity Tech: A Mission-Driven Career
Meet Rofy, a founder and technologist bridging the gap between cutting-edge AI development and social equity. From launching a successful beauty tech startup during the pandemic to advocating for stronger technical education, Rofy is proving that the data and the people behind it matter just as much as the code itself. Beyond his work as co-founder of yShade.AI, he is a dedicated mentor committed to equipping students with the practical skills to build real AI solutions. In this spotlight, he reflects on the evolution of his startup, the dangers of systemic bias in machine learning, and his efforts to empower the next generation of builders.
-Background & ESTEEM Experience-
ESTEEM: Can you share a bit about your professional journey after graduating from Notre Dame?
Rofy: After graduating from ESTEEM, I began my career as a technology consultant, starting out as a data engineer. The launch of ChatGPT in late November 2022 was a turning point for the industry, and as businesses began seeking more AI and generative AI solutions, I transitioned into a GenAI engineer role, advising clients and building solutions in the field. I also spent time doing traditional machine learning work, which remains relevant since not every problem requires the complexity of a large language model. From there, I moved into a GenAI Solutions Architect role, helping customers solve real pain points using third-party foundation models. Along the way, I had the opportunity to present at NVIDIA GTC 2024 and participate in the Google for Startups Program.
ESTEEM: Looking back, what part of the ESTEEM experience, whether a course, mentor, or project, had the biggest influence on how you work today?
Rofy: I loved the ESTEEM experience overall, from winning the intramural soccer competition with my cohort to winning the McCloskey competition with Chuqlab. It would be hard to single out one course or project, because the cumulative nature of the curriculum and my capstone work all pushed me to become a better entrepreneur. The biggest impact for me has been the overall approach to problem-solving. Having to conduct so many stakeholder interviews, organize and synthesize notes, and extract actionable insights trains you to be disciplined and patient. More importantly, it sharpens your ability to see the real pain points beneath what a customer says they need, which is a skill I use constantly in my work today.
-Career Evolution-
ESTEEM: You spent three years at EY as a Data & GenAI Engineer before moving to AWS as a Worldwide Solutions Architect. How did your experience building yShade prepare you and why did you transition from EY to AWS?
Rofy: yShade.AI was instrumental in preparing me for the work I do today. It was my baptism by fire, a period where I had to be heads down, figuring out the technology and building rapidly. I often say that yShade.AI gave me the builder's perspective on the GenAI stack. EY then gave me the consulting perspective, and moving to AWS gave me the cloud provider perspective. Each chapter was intentional, all in pursuit of becoming well-rounded and truly "full stack" in the field.
ESTEEM: At AWS, you recently architected SynapScribe, a voice-first Agentic AI platform. What was the biggest technical or strategic hurdle in completing that project and what skills from ESTEEM helped you navigate that challenge?
Rofy: The biggest technical challenge was attempting to use an any-to-any model (Qwen2.5-Omni-3B) for the complete ASR, Q&A, and TTS pipeline, given its ability to generate both text and natural speech responses in a streaming manner. Most architectures break each component of this pipeline into separate modules, each powered by a different, more specialized model. While modular, that approach can introduce unwanted latency if not carefully designed. Getting natural speech responses to work via vLLM in a self-managed deployment was particularly tedious, as native support for this was not available at the time of development. The ESTEEM skill that proved most applicable was rigorous documentation: the habit of thoroughly capturing findings to support recommendations and justify technical decisions.
-Building yShade.ai-
ESTEEM: Can you tell us about the origin of yShade.ai? How did you go from a concept to a product?
Rofy: The inception of yShade.AI, then called roboMUA, began in 2020 during the COVID-19 lockdowns. My co-founder and I were in New York at the time, and he had recognized a significant gap in how skin tones were represented in the beauty industry, something that had been on his radar since our college days. We started with just five skin shade categories, using celebrities as reference anchors to help us source publicly available data: Porcelain (Emma Stone, Ed Sheeran), Fair (Awkwafina), Medium (Kim Kardashian), Olive (Priyanka Chopra), and Dark (Lupita Nyong'o).
Because comprehensive datasets for diverse skin tones did not exist at the time, we scraped all of our own training data and built computer vision models from scratch. I handled full-stack development and designed the very first front-end. We launched across social platforms in March 2020, and by that summer, we had seen viral growth. Over time, we expanded to "yShade 100," adopting an inclusive 40-plus shade approach inspired by brands like Fenty Beauty to ensure true representation.
ESTEEM: Systemic bias is a massive hurdle in data and model training. How did you ensure yShade’s models remained representative where other platforms have failed?
Rofy: The core problem is that datasets lack depth for diverse populations because the teams driving these decisions are often not BIPOC, nor do they include BIPOC voices in the data collection process, which is a crucial stage that affects all downstream outcomes. We see this play out in the security sector through wrongful arrests and in the beauty industry through poor color matching. Bias gets baked into the models. Even in the generative AI era, we are still building on top of biased historical data. If the data is flawed, the solution will be discriminatory. We are trying to change that from the ground up.
ESTEEM: Often, the best co-founding teams have a mix of technical and commercial strengths. How did you divide the "business" versus "building" responsibilities at yShade?
Rofy: My co-founder and I were both technical founders, so we were intentional about building a team that covered the skillsets we lacked. We were fortunate to recruit colleagues from our college days, tap into the ESTEEM and broader Notre Dame network, and leverage LinkedIn to round out the team. Since we had the building side fully covered, our focus was on strengthening the business end. Bringing on team members for social media, graphic design, and advertising allowed us to divide the work effectively and avoid burning out faster than we otherwise would have.
-Advice for Students & Future Innovators-
ESTEEM: Many students are looking to enter the "hot" market of AI, what is your best advice?
Rofy: My advice is to shift your philosophy to: "Code is cheap, show me the work." In today's market, knowing how to use ChatGPT is not enough. You need a portfolio of three to five substantial, end-to-end AI applications that you have architected and built yourself, not superficial demos.
Pick a tool, develop deep competency, and build something for a niche use case. Whether you are aiming for a Solutions Architect role or a Go-To-Market Engineering position, you need concrete evidence that you can build. With tools like Cursor, Replit, and Claude Code, the technical barriers are lower than ever. There are no excuses. Just build.
ESTEEM: You’ve experienced the highs of creating a successful startup and the frustrations of navigating the tech job market. What is your takeaway for current students?
Rofy: My advice to students is to avoid getting stuck just checking boxes. The market has been tough, and it is important to invest in yourself beyond program requirements by building the technical skills the market actually demands. The resources to do this are largely free and widely available today.
Beyond technical skills, prioritize professional integrity, especially when networking. If you reach out to an alum for mentorship, follow through. If your direction changes, send a note. When you commit to something, see it through, and if you cannot, simply say so upfront. It is okay to say no. Relationships are hard to build and easy to damage, so protect them from the very start.