Assylnur Lesken

Data Scientist

DSRC

AI and data technology company

Fine-tunes language models and builds RAG and chatbot systems for legal document search, plus classification and price extraction scripts.

• Fine-tuned GPT-OSS-20B using LoRA for domain-specific legal QA. • Built a RAG AI Agent pipeline with Langgraph and Milvus/Postgres/Minio for semantic search over legal documents. • Developed a chatbot API with FastAPI and served the model via vLLM. • Created classification and price extraction scripts for product bills.

Data Scientist

Spark!

An innovation and experiential learning lab

Analyzed 70K+ legislative bills and built topic and network analysis; trained 97% accuracy literacy progression classifiers on data from 45K+ students.

• Scraped and analyzed 70K+ Massachusetts Legislature bills. Built a BERTopic and NetworkX (Louvain) pipeline to extract 20+ topics and map legislator networks, with 3× interpretability versus LDA. • Trained ML classifiers with 97% accuracy to predict literacy progression for 45K+ students using engineered survey and attendance features. Deployed insights in Looker Studio and Tableau dashboards.

Data Scientist

Egistic

Farm management system startup

Built a crop yield prediction model using satellite imagery and time-series data from 100K fields, with an MSE of 6.74.

• Trained a computer vision model for crop yield prediction using Sentinel-2 imagery and time-series data from 100K fields (MSE: 6.74). • Handled cloud-masked data and improved image resolution using Satlas preprocessing.

Summer Data Analyst Intern

PwC

Professional services firm

Built a CNN for livestock weight prediction and analyzed 500K+ retail products, reporting a 35% year-over-year price increase.

• Trained and deployed a multi-resolution grayscale-binary CNN for livestock weight prediction (MAPE: 0.18). • Analyzed 500K+ retail products and delivered a price trend report showing a 35% year-over-year increase for PwC Strategy& Kazakhstan.

Data Scientist Summer Intern

BI Group

Construction holding

Built an NLP model to forecast exchange rates and inflation from central bank reports, and developed geospatial property search.

• Built an NLP-based deep learning model to forecast the USD exchange rate and inflation from scraped central bank reports (R2: 0.70). • Developed natural language property search using H3 geospatial indexing for location-based recommendations.

M.S., Data Science

Boston University

B.Sc., Big Data Analysis

Astana IT University