m
mubtal

Dinmukhamed M.

@mubtal

AI, ML and Backend Engineer specializing in Python and LLM

Kazakhstan
English, Russian, Kazakh
About me
Hi, I'm Dinmukhamed, an AI/ML and backend engineer specializing in Python. I build custom AI chatbots (RAG/LLM), FastAPI backends, and computer vision models - production-ready code, not throwaway notebooks. My stack: OpenAI, Anthropic Claude, LangChain, PyTorch, PostgreSQL and Docker. I turn ideas into reliable, well-documented systems you fully own. Need an AI assistant trained on your data, a REST API, or a vision model? Message me before ordering and I'll scope the best approach for your project.... Read more

Skills

m
mubtal
Dinmukhamed M.
Offline • 

See my services

AI Chatbot Development
I will build a custom ai chatbot with rag
From$50 / projectMore details
Full Stack Web Applications
I will develop a full stack web app
From$120 / projectMore details

Portfolio

Work experience

Rogue_Realty

Backend / ML engineer

Rogue Realty • Full-time

Aug 2025 - Present • 11 mos

Designed and built the end-to-end ML pipeline in Python (PyTorch, NumPy) for analyzing JWST/NIRSpec spectra of rogue planets — raw spectra ingestion → preprocessing → classification Implemented the NIRSpec NRS1/NRS2 spectrum-stitching module — aligned and merged the two detector segments into a single continuous spectrum, handling the overlap region and gap edges Wrote the preprocessing stage: normalization, resampling onto a common wavelength grid, and noise handling, turning heterogeneous raw spectra into model-ready tensors Built a CNN classifier (PyTorch) for spectrum quality assessment and weak-signal detection, with a confidence-thresholded fail-safe that routes low-confidence samples to manual review Structured the code as decoupled modules (preprocessing / stitching / model / inference) behind stable interfaces, so the ML, frontend, and infra workstreams could integrate independently

Pexels

Backend Developer Intern

Pexels • Part-time

Jan 2026 - Apr 2026 • 3 mos

Built a FastAPI endpoint with Pydantic validation and PostgreSQL integration — shipped in the first 2 weeks Migrated session auth to JWT + refresh rotation, fixing 3 recurring bugs and cutting login latency by ~40 ms Traced an N+1 query in the user-listing endpoint; eager loading dropped DB round-trips from ~80 to 3 per request Wrote pytest integration tests for 8 critical routes against a live test container, catching 2 regressions pre-staging Added pre-commit hooks (Black, isort, flake8) - formatting comments in PR reviews dropped by ~60%