Moscow Institute of Physics and Technology (State University) (MIPT)

Head of Labeling at Dbrain

Revolut

Product Manager, AI

Everinfer is a low-latency serverless AI platform. 
As a product co-founder, I defined the requirements for customer-facing SDK, implemented demos and some AWS infra code, and acquired initial customers and testers.
Failed to get enough velocity, shut down.

Everinfer

Co-Founder

I work with ML devs and analysts to come up with solutions to drive the efficiency of our dispatching by building algorithms to compute optimal routes, predict supply/demand, and assign batched orders to best-fitting couriers.


Our team re-designed the whole platform within two years to support explosive growth from 200k to 1+kk deliveries per day, increasing batch rates up to 40+% and increasing savings due to our algorithms by up to six figures in $ per day.

Yandex Delivery

Technical Product Manager

At Lamoda, the most technologically advanced company of Global Fashion Group, I, as an individual contributor: 

- Redesigned cross-sales recommendation system from the ground up by implementing custom matrix factorization algorithms. Installed a process to collaborate with the product management tea, to mine new insights on UI composition from data.  As a result, cart adds from recommendations were increased by 59% since Sep. 2019. 

- Developed the learning-to-rank core for a novel catalog ranking system. Redesign of catalog ranking done in 2021 brought in more than +1% NMV.

As a manager, I:

- Was tightly involved in the hiring process, streamlined tech interview structure, and drove company brand through delivering inspiring presentations on team mission and tech to candidates. 

- Sourced and managed external partners to reinforce our computer vision capabilities through outsourcing

- Managed tech communications with parent companies and other holding subsidiaries, and delivered presentations on the state of our data-driven products to external stakeholders.

Lamoda Group

Senior Data Scientist / Data Product Manager

NeuroCore is a Computer Vision and Machine Learning consulting team, specializing in full-cycle projects.
As a managing partner, I drive team growth through:

- Establishing partnerships, securing long-term contracts with new clients, and expanding into the European market.

- Hiring (6 -> 13 full-time developers)

-Pre-sales activities and pitches, transforming client-side business requirements into clearly defined technical tasks.

Neurocore R&D Lab

Co-founder, Head of AI

As a Head of Labeling at Dbrain, I managed the development and sales of a crowdsourcing platform to annotate and manage data for CV & NLP.

 - I structured sales activities of a team of 4 people and led product pitches. Onboarded clients include, but are not limited to Amazon, Nvidia, Tinkoff, MTS AI, Cherry Labs.

 - Secured partnerships with BPOs that significantly improved our capability to annotate sensitive data and became the first certified partner of Y.Toloka.

 - Led the development team on a mission to implement the best data annotation tool on the market.

Handl.ai

Head of Data Annotation Platform

As a mentor on Data Science course i consulted multiple students in building computer vision systems from scratch for their final projects.

Нетология-групп

Mentor

I took part in all stages of multiple consulting projects, from pre-sales to production. 

Notable projects:
 - 4k-resolution image segmentation for retail photography
 - Aesthetic photo enhancement tool that mimics the workflow of a professional photographer
 - Livestock detection and recognition engine for CattleCare

The last one even got featured at TechCrunch

Senior Data Scientist

At MTS, the largest mobile operator of Russia and CIS, i was entrusted to lead the development of a “credit-scoring-as-a-service” product that predicts credit scores for subscribers of other operators on a daily basis. Predictions of that service are used at multiple banks. 

Other projects that i worked on include a smartphone recommender system for e-commerce department, extracting interpretable user interests description based on analysis of mobile web traffic and an internship program

MTS Group

Data Scientist

At Connectome.ai i led the development of an MVP of computer vision system to estimate quality of processed meat on conveyor lines.

Connectome.ai

At Habidatum i helped policymakers and businesses to understand the urban environment through geo-spatial data analysis.

Besides ad-hoc analytics projects I built a fully-automated text sentiment classification engine that works on any language. It scrapes training data from the geo-tagged social networks and selects optimal neural network architecture automatically. 

Take a look at the results, visualized through Habidatum’s excellent Chronotope tool:

Habidatum

I did multiple part-time jobs, notable ones being a system to scrape data from the web about people that pose financial crime risk (ComplyAdvantage), and a recommender system for e-commerce startup that sells wine (Invisible).

Andrey Kiselev