Daniiar Abdiev

Aiding an ed-tech startup to develop an AI chatbot for quiz creation and checking using OpenAI, Pinecone, and Langchain.
Consulted on a Telegram-based LLM AI therapy chatbot project with AI voice generation.
Advised on an LLM AI project for SQL database queries beyond AI's context size, using Langchain and Pinecone.

• Led the development of ChatGPT like Chatbot ordering experience for grocery. Featured in PCMag https://shorturl.at/fOW14
• Led the development of ChatGPT integrated Support Chatbot reducing 8% percent of contact rates with just a simple MVP.
• Developed a semantic search engine for food and grocery items using LLM, GPTCache, with data stored in a pg vector store.
• Developed a semantic similarity algorithm for grocery items, leveraging NLP and CV, generating a $4M business opportunity.
• Led the development of an out-of-stock detection model, resulting in $300K GMV savings in three months.

Aiding an ed-tech startup to develop an AI chatbot for quiz creation and checking using OpenAI, Pinecone, and Langchain.
Consulted on a Telegram-based LLM AI therapy chatbot project with AI voice generation.
Advised on an LLM AI project for SQL database queries beyond AI's context size, using Langchain and Pinecone.

For Sony LATAM, Developed an algorithm for predicting the number of future song streams, enabling the company to optimize its marketing and promotion efforts for maximum impact.
Forecasted streams for more than 200 artists.
Developed an algorithm for predicting the admission possibility of students, improving the accuracy and efficiency of the university admission process.
For ForeSquire, Developed a geo data algorithm for forecasting the number of customers crossing certain intersections, enabling the company to optimize their marketing and advertising strategies based on foot traffic data, resulting in over 1M in ad sales.
Led the development of a SaaS product that created data dashboards based on social media data per user.

Led the development of an ensemble of neural network and extreme gradient boosting models for gender and age classification of users using calling metadata.
Achieved state-of-the-art accuracy of close to 80%.
Developed an ensemble of extreme gradient boosting models for primary-secondary SIM card classification, using SHAP for interpretation.
Increased accuracy by 20% and automated the entire pipeline.

For Sony LATAM, Developed an algorithm for predicting the number of future song streams, enabling the company to optimize its marketing and promotion efforts for maximum impact. Forecasted streams for more than 200 artists.
Developed an algorithm for predicting the admission possibility of students, improving the accuracy and efficiency of the university admission process.
For ForeSquire, Developed a geo data algorithm for forecasting the number of customers crossing certain intersections, enabling the company to optimize their marketing and advertising strategies based on foot traffic data, resulting in over 1M in ad sales.
Led the development of a SaaS product that created data dashboards based on social media data per user.

Led the development of an ensemble of neural network and extreme gradient boosting models for gender and age classification of users using calling metadata. Achieved state-of-the-art accuracy of close to 80%.
Developed an ensemble of extreme gradient boosting models for primary-secondary SIM card classification, using SHAP for interpretation. Increased accuracy by 20% and automated the entire pipeline.

Led the development of ChatGPT like Chatbot ordering experience for grocery.
Led the development of ChatGPT integrated Support Chatbot reducing 8% percent of contact rates with just a simple MVP.
Developed a semantic search engine for food and grocery items using LLM, GPTCache, with data stored in a pg vector store.
Developed a semantic similarity algorithm for grocery items, leveraging NLP and CV, generating a $4M business opportunity.
Led the development of an out-of-stock detection model, resulting in $300K GMV savings in three months.