Andrew Stepanov

I am an engineering manager with first-hand experience in the field of machine learning and data analysis, including all stages of the ML pipeline from setting the objective to production. I know how to build and manage teams making services that sustain tens of thousands of requests per second under load running ML-heavy algorithms. Being a person of technical and ACM background, I have a deep understanding of how various aspects of modern ML in production stack up, including cold starting, perf optimization (CPU/GPU), cloud-native architecture and distributed systems, quality monitoring and continuously improving the model, etc.

Areas of interest and expertise: machine learning, distributed systems, cloud-native architecture, speech recognition and synthesis, C++ and Golang development, IT process improvement, big data processing.

Work Experience 💼

Tinkoff

You can view recommendations from my peers and superiors on LinkedIn.
  • Head of Department, ML Infrastructure and Services

    -

    I manage a team of ~60 people with a primary goal of developing and maintaining a large suite of internal instruments, infrastructure, and services aimed at reducing time to market and increasing ease of use for machine learning-based workflows and products at the company. This includes computer vision and document recognition services, speech services, training and serving infrastructure, confidential computing, etc. As a result, most ML engineers in the company do it via tools developed in my department and a significant portion of the company's business teams use our ML SaaS solutions in their critical business paths.

  • Head of Tinkoff VoiceKit

    -

    My team and I set out to test if a new SaaS business can be launched based on the voice tech we developed in the years prior. My direct reports and I were responsible for go-to-market strategy, product offering, engineering management, and sales. We managed to get 10s of paying enterprise customers and met our projections in the first and second years we operated. Afterward, we deemed our idea successful and the business part transitioned to the separate IT monetization department.

  • Lead Data Scientist / Team Lead

    -

    I managed a team of a dozen engineers and also took an active part in developing automatic speech recognition and speech synthesis systems from scratch for one of the largest fintechs in Europe. My team was responsible for all parts of the machine learning cycle - from data collection and labeling to production deployment and quality monitoring. The system that we've built now processes tens of billions of seconds of real-time speech, each month.

Vivid Money

  • Machine Learning Consultant

    -

    I consulted one of the biggest German fintechs in applying Machine Learning for customer care and support. As a result of this work, the company managed to build an in-house team with the required expertise and focus on that particular area.

Education 🎓

Moscow Institute of Physics and Technology (MIPT)

  • Master of Science, Applied mathematics and Computer Science

    -

    Specialized in data analysis and machine learning. The thesis topic is Sequence discriminative training and structured prediction for end-to-end speech recognition.
    GPA: 4.6 / 5

  • Bachelor of Science, magna cum laude, Applied mathematics and Computer Science

    -

    Specialized in computer science and machine learning. The thesis topic is Character-level RNNLM for morphologically complex languages in speech recognition.
    GPA: 4.9 / 5

Yandex School of Data Analysis

  • MSc-equivalent degree, CS department

    -

    The studies were mostly independent from the main ones in MIPT. During my studies, I took a deep dive into reinforcement learning, applied statistics, distributed systems and concurrency, computer vision, and machine translation.

Summer Computer Science Camp

  • Student, A' stream

    I took this summer camp to better prepare for upcoming competitive programming contests. I intensively studied advanced algorithms and data structures during the 3 weeks of the summer of 2012. As a result of this study, I won the regional stage and was awarded a prize at the national stage of the Russian Informatics Olympiad (ROI) for high-school students.

  • Student, B stream

    I first took Summer CS Camp to hone my skills in algorithms and data structures during the summer of 2011.

Samara Lyceum of IT

  • Student, magna cum laude

    -

    The lyceum focuses on the in-depth study of information technology. I started programming in 5th grade. During my high-school years, I specialized in physics and mathematics.
    High-school GPA: 5.0/5