Kim Scott, PhD

Fractional Chief Data Scientist, AI Product Manager, Entrepreneur

I’m a seasoned data science leader with over 15 years of experience developing cutting-edge machine learning and AI solutions. After spending eight years in executive roles at early-stage startups, I have become skilled at crafting and executing strategic directions for technical teams, navigating fast-paced environments that require quick adjustments and resourcefulness with ease. I am recognized for bringing clarity to chaos by aligning business and technical requirements, demonstrating empathetic people leadership, and fostering cross-functional collaboration. I’m proud to drive the development of impactful ML/AI solutions and products, leading teams of up to 12 talented professionals while managing budgets ranging from $1M to $8M. My focus is on achieving business goals and ensuring customer satisfaction, empowering me to leverage my expertise in startups to hit critical technical milestones efficiently and with quality outcomes.

Career Accomplishments

Securing Revenue from NASA: A Collaborative Success at Pixxel

At Pixxel, a young startup in the Earth imaging sector seeking to sell data to US federal agencies, we faced a significant challenge when responding to a critical NASA proposal. After two weeks of stalled progress from our technical team, I directed the writing effort for our technical volume. I clarified roles, implemented an asynchronous communication plan, and actively contributed to the proposal, leading to an on-time submission. Six months later, NASA chose us as one of eight companies for a $476M multi-vendor award and secured our first order for $1.2M in imagery sales, marking a significant milestone for Pixxel.

Transforming Soil Sample Management with the SoilCollector App

At EarthOptics, we faced significant challenges with inaccurate labeling and geolocation of soil samples, leading to wasted time and increased errors. Tasked with addressing this issue, I led a team of three software engineers in designing and deploying a new offline application to streamline soil sample data collection. By coordinating with different teams to map our workflow and identify bottlenecks, we created a solution that improved operational efficiency and reduced data errors by 20%. This application simplified processes across the organization and fostered better collaboration among departments.

Developing the C-Mapper ML Solution: Scaling for Success

Within just four weeks of starting my role at EarthOptics, I prototyped an ML solution using ground-based sensor data that garnered significant interest from our earliest customers. We encountered one of those “great problems to have” moments just four months later when we realized we had lined up ten times the demand for the Fall season. However, we faced the challenge of lacking an automated data processing pipeline and relying on cumbersome code for one-off solutions. To address this, I advocated for a shift in priorities to our leadership to focus on building an automated ML pipeline for four weeks. With their approval, I coordinated with my data science team to clear their schedules, enabling us to deliver the automated ML pipeline in time for the new season’s data. This greatly reduced customer delivery timelines by 80% and enhanced operational efficiency.

Grant Award to Demonstrate AI’s Value in Climate Solutions

After closing our Series A funding round at Astraea, I was introduced to the program office of our lead investor to explore partnership opportunities that would showcase our platform’s scalable AI capabilities using Earth Observation data. As the technical lead, I organized and facilitated technical exchange meetings between our cross-functional teams and potential partners to identify ambitious ML/AI projects that addressed real-world climate challenges. Over six weeks, I collaborated with other companies to pinpoint an impactful climate-related problem, outline a solution utilizing satellite imagery and deep learning, and establish a comprehensive scope of work that aligned with Astraea’s business objectives. This concerted effort resulted in securing a $420k grant to develop a solution that generates a global database characterizing GHG emissions from undocumented cement and steel industries, marking a critical milestone by generating our first revenue from a paying customer.

Successful Development and Delivery of a Deep-learning Model at Astraea

When my startup’s first paid contract involved partnering with two institutions to train and deploy a large-scale CNN model using Earth Observation data, I managed customer and partner relationships to ensure success. Initially, our partners were responsible for designing and training the model, while my data science team focused on deployment. However, after identifying critical data preparation issues that would affect our model’s performance, I stepped in to redesign the training process and build the deployment pipeline myself. My proactive approach and commitment to quality allowed us to deliver high-quality results, generating our first revenue and establishing a high-impact ML solution that was later published in a peer-reviewed journal.

Launching the Astraea EarthAI MVP: Turning Vision into Reality

In an “all hands on deck” situation at my seed-stage startup, we needed to design, develop, and launch a minimum viable product (MVP) for EarthAI to prepare for a live demo with an interested investor. With no existing product, product manager, or development processes and a tight timeline of six weeks, I stepped up to lead the initiative, coordinating all technical resources, including three data scientists and five software engineers. I defined and adapted product requirements to align with our demo goals, established a cross-functional agile development process, and guided the team in UI design and product testing. We successfully launched the MVP as planned, impressing the investor during the demo, which extended twice as long due to heightened interest and ultimately contributed to our successful Series A funding round.