Dmitrii Lapin

linkedin // @lapin_dv // @digitaloutsider

<aside> <img src="/icons/mountains_gray.svg" alt="/icons/mountains_gray.svg" width="40px" /> I’m Head of Analytics / AI Product Lead with 10+ years of experience building and scaling data-intensive products at companies like Yandex and Avito. Delivered AI/ML solutions generating over $30M in annual revenue, from initial concept to market launch. Proven ability to build and lead cross-functional teams (30+ specialists) and define product strategy from 0-to-1 in both corporate and startup environments

</aside>

<aside> <img src="/icons/briefcase_gray.svg" alt="/icons/briefcase_gray.svg" width="40px" />

Lead DS Product Manager (Jan 2025 – Dec 2025) – Avito (top-1 World classified from Russia)

Responsible for planning and launching ML-based products within the new Avito Ads platform (ads.avito.com), as well as for developing inventory monetization and new ad placements across all verticals

• Delivered the MVP for the new Avito Ads ML platform in one quarter, including a CTR-prediction model for ad placement, and defined the product roadmap for Q3 2025.
• Generated an estimated +$5M in annual revenue by personally defining and validating hypotheses for the new ad platform.
• Developed the 2026 product roadmap with a pipeline of features projected to add an additional +$35M in revenue.
• Created and led an internal masterclass for Product Managers on applying LLMs, RAG, and deep search techniques to daily workflows, improving team-wide efficiency

Head of AI and Analytics (Aug 2024 – Dec 2025), part-time – RetailHub (the coolest retail startup in UAE)

Invited to build the AI technology stack and products from scratch (0 → 1) in a startup focused on digitalizing physical stores through shelf scanning and product matching across various tasks and processes

• Independently developed a Python-based product-matching service using a cascading architecture (TF-IDF + text embeddings → LLM validation → fine-tuned Electra binary classifier). Reduced matching time from 4–5 hours to 30 minutes with high recall and precision (>0.8)
• Led the AI development of the flagship computer vision shelf-scanning product, improving product-matching accuracy from 65% to 95%+
• Built a GPT(VLM)-based assessor tool, cutting manual product acceptance work in half and saving ~1000 $ per store
• Set up and maintained analytics for the flagship CV product using Supabase and Preset (cloud Superset). Built a system of 3 main dashboards: from store business metrics (AOV, ABV, OOS, SKU count) to accuracy and performance metrics (confidence level per product, processing time per photo, number of products per photo)

AI Expert (Sept 2024 – Feb 2025), part-time – FORTIS (another cool POS startup in UAE)

Initially the second person responsible for AI development in the company; after two months became the sole lead of the AI track

• Independently launched 10 AI PoCs in one quarter (Python + APIs on Streamlit), resulting in several successful presales and integration of 2 features into core products:
  • smart table import
  • automated text generation for products and services
• Built several internal AI tools using low-code n8n to increase efficiency:
  • advisory bot for sales managers (reduced product info search time from up to 30 minutes to 1 minute)
  • web interface for uploading and quickly analyzing qualitative & quantitative UX survey results
  • automatic call transcription & summarization (RU/EN/AR mixed) with CRM task generation
• Within one month, developed a knowledge-base-connected chatbot for C-level executives with features such as:
  • task creation in trackers & calendar integration
  • transcription of voice, video, and various document formats
  • RAG support with wiki & task tracker integration
  • web search & deep research

Professor (Jun 2024 – Sept 2024), part-time – IT Career Hub (German education hub)

Delivered lectures for beginner and intermediate students on product analytics

• Prepared complete course materials on product analytics “from scratch” (covering HADI cycles, A/B/n testing, etc.) — 9 lectures, each 4 academic hours
• Successfully trained a group of 15 students through a 3-month program

Product Lead (Jun 2023 – Jun 2024) – Yandex (Russian bigtech)

Initially invited to establish product-market fit for a warehouse robotics prototype. Eventually pivoted the project based on new design and analytical modeling; the new simplified robots are now successfully deployed for order fulfillment: Yandex Robotics Case

• Identified product-market fit, north-star metric, and business impact within the first two months; consolidated into a single business model and demonstrated inefficiency of the earlier concept
• Conducted market research and, with an analyst, developed simulation models for various robot & warehouse configurations — selected the optimal solution for constraints and formulated a technical specification for engineering
• Built a detailed business model linking robot dimensions → estimated darkstore GMV impact via storage capacity & AOV growth through OSA improvements
• Diversified the earlier prototype toward more relevant use cases and successfully completed a presale for vertical farms (although later discontinued due to market reasons)
• After internal transition, initiated a stream of LLM-feature integrations into e-grocery software products (e.g., PIM content management, WMS acceptance, etc.)

Head of DA/DS and CBDO (May 2021 – April 2023) – JSA Group (big IT company)

Invited with the team to establish industrial-grade data analytics from scratch and rapidly deliver a critical flagship project in safety video analytics for manufacturing plants. Then brief but effective tenure as Business Development Director, focusing on product systematization

• Within two months, together with DS and MLE team, developed an MVP safety video analytics solution and successfully launched it on 20 cameras across 2 plants (press release). Six months later scaled to 200+ cameras across 4 plants (JSA Sees), closing multiple workplace injury prevention cases
• Designed system architecture and concepts for SOTA AI products:
  • shop-floor safety video analytics
  • granulometric composition and conveyor belt condition analysis
  • railway collision prevention system (link)
  • driver monitoring for trains and dump trucks (press release)
• These projects generated a total annual revenue of 5M $
• Implemented operations, economic, and technical analytics processes: identified dangerous zones for excavator buckets (frequent tooth breakages), determined best equipment suppliers, eliminated procurement fraud in PPE, and optimized sizing grids — reducing clients’ operational costs by 2,5M $ annually
• Structured 21 products out of 48 projects based on TRL and market research
• Created a sales kit from these 21 products, later applied successfully in sales; 11 products were listed in the Russian software registry (solutions catalog)
• Independently conducted presales, successfully advancing 2 products to deals worth 1M $

Lead of R&D Laboratory (May 2019 – April 2021) – Ctrl2Go (top-5 Integrator in heavy industry and my first AI startup)

Initially hired as a product manager’s assistant, later advanced to lead an R&D division focused on AI applications in sensor data analytics. Developed a “Shazam” for equipment sound diagnostics

• Together with one developer, created a hardware-software system for acoustic diagnostics using heuristic sound pattern recognition — successfully lab-tested and piloted at a locomotive factory (paper abstract)
• In parallel, expanded the team and closed an R&D project on acoustic-emission analysis of oil & gas pipelines with automatic ML-based diagnostics (paper abstract)
• During COVID-19 isolation, enhanced the system with unsupervised learning and anomaly detection on encoders/decoders; later patented the solution (Ctrl2GO patent)
• Built and maintained the company’s R&D matrix, systematized products, and advised on acquisition/investment in 5 industrial deep-tech startups
• Independently hired and formed a team of 5 top specialists in data analysis and software engineering
• Successfully closed projects and studies with a total budget of ~1M $

Researcher (2016 – 2019) – Bauman Moscow State Technical University

Worked in a micro-group at a research institute.

• Built mobile robots, developed algorithms, worked with ML/AI (before it was widely called that). Publications available: Google Scholar profile.
• Successfully completed 7 R&D projects with a total budget of ~10M $ </aside>

<aside> <img src="/icons/sunglasses_gray.svg" alt="/icons/sunglasses_gray.svg" width="40px" /> Challenges

• 1st place in Volga Iron Hack, 3rd place in EVRAZ AI Challenge, solo 4th place in BIG GEN AI Hackathon
• Demo of AI-first game for jam Vas3k.Club Jam V, won 1st place and got cool honest review

Mentor in DA/DS: 30+ mentee, helping with training in DA, DA/DS case solving on Solvery, GetMentor

Ex-Scientist in Robotics and AI: 34 papers, 20 in SCOPUS / WoS, 2 in Q2/Q3

</aside>

<aside> <img src="/icons/graduate_gray.svg" alt="/icons/graduate_gray.svg" width="40px" /> Bauman Moscow State Technical University

Master (2010 – 2016) – GPA: 4.9 / 5

PhD (2016 – 2020) – GPA: 4.5 / 5

</aside>

<aside> <img src="/icons/asterisk_gray.svg" alt="/icons/asterisk_gray.svg" width="40px" /> Languages: Russian (native), English (B2)

Stack: Product Strategy / AI/ML Product Development / Roadmapping / Python / SQL / Computer Vision / LLMs / RAG / Data Analytics / Superset / A/B Testing / System Design

Hobbies: 3D-printing, boxing, enduro, DIY for home, whiteboards

</aside>