Open to opportunities

Kudaibergen Abutalip

@kudaibergenabutalip

Message

I build production ML and AI systems for computer vision and operations.

Kazakhstan

Message

What I'm looking for

I'm seeking Senior ML/AI Engineer or Senior SWE/ML Engineer roles where I can build and scale production machine learning, computer vision, LLM, and MLOps systems.

At Aldente AI, I own production AI systems for restaurant operations, from asynchronous order verification to LLM-powered dispute automation and tool-using RAG agents.

I designed a continual-learning pipeline that converts production traces into retraining data, reducing false negatives from 15% to 3%. I also tuned SGLang serving and batching to scale from one to 11 restaurant locations, and built dispute automation that achieved a 75% hit rate while reducing inference costs by 80%.

Previously at VEON, I led development of an OCR engine for a biometric identity system handling 20K+ daily verification sessions, reducing end-to-end latency by 25%. I mentored two junior data scientists and helped lead sprint planning for a five-person ML team.

At TargetAI, I built asynchronous inference pipelines for real-time video analytics across 30K+ cameras and improved detection, recognition, and deployment workflows. My work reduced false positives by 80%, raised CRNN accuracy from 90.2% to 98.3%, and cut latency by 20x.

Experience

Work history, roles, and key accomplishments

Current

Founding AI Engineer

Current

Aldente AI

May 2025 - Present (1 year 3 months)

Owned the production order-verification system, architecting asynchronous inference with config-driven routing across self-hosted SGLang and cloud providers. Designed the continual-learning pipeline that turned production traces into retraining data, cutting false negatives from 15% to 3%.

Python PyTorch Sglang LiteLLM fastAPI Langfuse OpenTelemetry Docker Kubernetes PostgreSQL Vertex AI OpenAI Claude Bedrock MiniCPM LLaMA Factory MCP RAG

Machine Learning Engineer

VEON

Sep 2024 - May 2025 (8 months)

Led development of the production OCR engine powering a biometric identity system processing 20K+ daily verification sessions, reducing end-to-end latency by 25%. Researched and built generative CV demos including style-preserving image synthesis and avatar generation using diffusion models.

Python PyTorch OCR Computer Vision Diffusion Models MLOps Sprint Planning

Computer Vision Engineer

TargetAI

Feb 2023 - Sep 2024 (1 year 7 months)

Developed teye-va (C++, Lua), a core module for real-time video analytics across 30K+ cameras by implementing asynchronous ML inference pipelines. Designed an end-to-end MLOps workflow for YOLOv8, reducing false positives by 80%, boosting CRNN accuracy from 90.2% to 98.3%, and cutting latency by 20x.

C++Lua YOLOv8 MLOps CI CD Computer Vision RAG PaliGemma