MVP Lab

Order a project

We design and develop new AI products and
AI functionality for existing products

PACKAGE YOUR EXPERTISE INTO AI PRODUCT

COMPLETE YOUR PRODUCT WITH AI-FUNCTIONALITY

OPTIMIZE YOUR PROCESS WITH AI

WE WORK WITH YOUR CONTEXT

Need a quick result

We understand the entrepreneur's vision and deliver an applicable result within the fixed time frame.

No data for AI product development

We collect data, use AI algorithms to start from scratch. The problem of small data is on us.

Flexibility and transparency

We make short iterations with demo of the result. Feedback affects development process.

It's hard to find a trusted AI specialist

We send the final result to the client. Finding a team for development does not bother the client.

OUR CASES

Photo Tags for medical test

Result: AI-functionality of the clinic's mobile application. The model puts the test category on the photo of printed doc.
Work: collected a sample of tests and markup, trained a type recognition model, developed a web service via REST API.
Client: MTS, Medsi.

Search for duplicate products

Result: optimized the process of matching product cards (jewelry) between supplier catalogs and the reference catalog.
Work: trained the embedding model and the proximity function of two products, developed a web-service.
Client: Gems

Content Recommendation

Result: AI-functionality of the bank's mobile application was created. The model recommends content based on customer transactional activity.
Work: Segmented the client base with a topic model, interpreted the results for a bank analyst.
Client: Bank (under NDA).

Price Prediction

Result: We optimized the process of purchasing steel products by predicting possible price levels in the future.
Work: Conducted an audit of the current process and model, expanded the factors, improved the model, developed a web service.
Client: Developer (under NDA).

Tender Documentation

Result: We optimized the process of selecting fields and combining information about tenders from websites of different companies and from different countries.
Work: Audited the parsing process, automated the addition of a new source, created NER models, etc.
Client: CubeRM.

Ring Size Estimation

Result: AI functionality for the website of the pawnshop network. The model for the photo of the hand estimates the size of the ring for each finger.
Work: Built a model of a skeletal image, segmentation of the hand, size estimation and developed a web service.
Client: Malachite.

Book a Team for Your Project

TECHNOLOGIES THAT WE USE

Language: Python
Backend: Python Flask, Django, REST API, etc.
Frontend: React, vue.js, etc.
Databases: PostgresSQL, MongoDB, etc.
Notebooks: Colab, Jupyter, etc.
Models Packaging: Kubernetes, Docker, Airflow, etc.

NN Frameworks: PyTorch (прототипирование), TensorFlow (production), TensorFlow lite (импорт на устройства);
Data Analytics: pandas, numpy;
Specialized Frameworks: BigARTM, TopicNet, gensim, nltk, DeepPavlov, SpaCy, OpenCV, scipy, etc;
Experiments Tracking: Wandb, MLFlow, Tensorboard;

Codebase Forming:

Read Me in repository;
Code Style: pep8;
High level of code readability.

Delivery Variants:

Docker + REST API;
Web-service + Frontend;
Git-repository: Python-scripts or library;
Web-site, mobile app in App Store / Play Market, etc.

PRICING

FIXED
PRICE

Work: we fix the SoW for the stage, flexibly change between iterations
Acceptance: according to the scenario and the criteria specified in SoW
Payment: advance payment and stage payment

from 6 to 40

kUSD: per project stage

Time
materials

Work: time is paid, works are agreed for each iteration
Acceptance: at the end of the iteration (2 weeks), the next iteration is approved
Payment: at the end of each iteration

from 50 to 70

USD: per work hour

Book a Team for Your Project

HOW WE WORK TOGETHER

1

Find out the requirements and context, sign the NDA (1 Day)

2

We evaluate the project and draw up a preliminary plan (3 Days)

3

We will draw up a SoW and an Agreement, sign it on both sides (2 Days)

4

We will make the first prototype and clarify the requirements (from 1 to 3 Weeks)

5

Development by iterations and delivery to the client (2 Weeks in 1 Sprint)

6

Handover and acceptance of the project stage (from 2 to 4 Sprints)