Abdirakhman Ismail

Backend and GenAI Engineer

Nace AI Kazakhstan

Builds backend and GenAI features for Nace Data Intelligence, improving document extraction and agent quality while developing evaluation, observability, and platform tooling.

• Improved OCR/document parsing accuracy by 23–48% and grounding accuracy by 11–20% across document benchmarks by optimizing diagram-parsing and grounding algorithms; led creation of OCR parsing ground truth via annotation of 1,000 PDF pages on customer-hosted infrastructure. • Built comprehensive finance-domain evaluation datasets spanning long-form documents, images, and structured tables, collecting 300+ financial files/images and 700+ table queries and generating 900+ QA pairs to benchmark file-search and multimodal query performance.

Founding Software Engineer

WWBW

Led backend development and Azure infrastructure at a startup, building an LLM ordering agent and product infrastructure used in raising $600,000.

• Contributed to securing $600,000 in investment by building and demonstrating robust, scalable product infrastructure and AI-driven features. • Architected and managed Microsoft Azure cloud infrastructure independently, establishing automated CI/CD deployment pipelines with GitHub Actions and secure environment configurations. • Built an LLM-powered agent that automates customer orders from natural language and image input, improving user convenience and conversion. • Led full backend development for the startup, including API design, database architecture, and deployment processes.

Software Engineer Intern

Big tech redacted

Added customer data export functionality and improved the resilience of data transfer pipelines during a software engineering internship.

• Added a long-awaited feature for exporting customer data from [Big tech] servers to user PCs. • Improved the resilience of [Big tech] customer data egress and ingress pipelines by implementing adjudication logic for a database plugin dependency.

Research Intern

KAIST SoftSec Lab

Built components for a binary analyzer, adding RISC-V support and cutting binary emulator runtime by 20%.

• Contributed to a binary analyzer project by implementing a parser, binary lifter, and emulator, enabling RISC-V support and optimizing x86-64 emulator performance. • Reduced binary-emulator runtime by 20% by implementing dead code elimination and constant folding.

Software Engineer Intern

Noteing

Reduced annotation data storage by over 60%, roughly halving database costs, and built payment and login flows for mobile apps.

• Reduced annotation data size by more than 60% through research into data-reduction techniques, cutting database server costs roughly in half. • Implemented payment and login pages using Flutter, and engineered a hybrid Flutter/Swift application to enable a smooth migration to both Android and iOS.

Bachelor of Science, Computer Science

Korea Advanced Institute of Science and Technology (KAIST)