h
hurualqayeva

Huru Algayeva

@hurualqayeva

AI Solutions Developer

Azerbeidzjan
Engels, Turks, Azerbeidzjaans
Sommige informatie wordt in het Engels weergegeven.
Over mij
AI Solutions Developer with 5+ years in LLMs, ML and data. I build production AI end to end: multi-agent workflows (LangGraph, LangChain), RAG chatbots, MCP servers, LLM integrations, LoRA fine-tuning and predictive models, with Python/FastAPI backends, React frontends and Docker/Kubernetes deployment. Recent work: a multi-agent system used by telecom operators, a finance reporting agent at KPMG (92% accuracy on 250 real questions), and a budgeting app that cut a 2-month process to minutes. Tell me your problem and I'll build the simplest AI that solves it!... Lees meer

Skills

h
hurualqayeva
Huru Algayeva
offline • 
Gemiddelde reactietijd: 1 uur

Bekijk mijn diensten

AI-integraties
I will build a custom ai agent, rag chatbot or llm app for your business
BI-analyse
I will build an ai data analyst that answers questions from your database

Werkervaring

Nasc_Technologies

AI Solutions Developer

Nasc Technologies

Dec 2025 - Present • 11 mos

Contract role on the development team building production AI agents. - Contributed to a production Python SDK for AI agents (6 releases): RAG integration, native MCP client support, lifecycle hooks, structured Grafana logging and developer docs. - Integrated LiteLLM as a multi-provider gateway for 10+ production agents, with fallback routing and per-call cost tracking. - Owned LLM token tracking and an admin dashboard end to end across 4 codebases (SDK capture, API, aggregation endpoints, React frontend). - Designed and shipped 3 LLM analysis scenarios in a LangGraph multi-agent system for telecom asset management, now used by telecom operator clients. - Built multi-agent workflows with LangChain/LangGraph, long-term memory and RAG (ChromaDB, pgvector, sentence-transformers); removed ~1,800 lines of duplicated code by moving shared data-layer logic into the SDK. - Fine-tuned LLMs with LoRA (Hugging Face, PEFT, Unsloth) for 2 domain-specific targets. - Raised mutation-test kill rate from 58% to 95.8% and fixed 10+ security vulnerabilities in internal AI/MCP services. - Deployed services with Kubernetes, Helm, GitLab CI/CD and S3-compatible storage; wrote RFCs and ADRs that shaped the team's AI roadmap.

KPMG

AI Engineer

KPMG • Fulltime

Dec 2024 - Dec 2025 • 1 yr

AI Engineer in the Technology Department, promoted within 10 months and given the Bravo Award for data solution design. - Architected a unified reporting platform for a holding of 30+ companies (retail, cinema, logistics, chemicals, energy). - Designed and shipped a production LLM reporting agent used by the client's finance teams: natural-language questions are grounded to the warehouse schema, compiled to ClickHouse SQL and answered with figures and variance explanations. Validated on 250 real analyst questions (92% correct) before release. - Designed hybrid data architecture across 20+ client projects: AWS (S3, RDS/Redshift) and on-prem sources landed into ClickHouse with Airflow and dbt, with IAM access and cost/retention policies. - Worked directly with finance, sales and C-level stakeholders to gather requirements, define KPIs and model the data (fact and dimension tables, grain, keys, naming conventions). - Wrote the technical proposal and solution design for a Real Estate digital twin, approved by the client, targeting a 35% cost reduction. - Fixed production DAX errors behind KPI mismatches, added validation checks and Python automations. - Ran code reviews, onboarded 12 new joiners and planned team tasks.

ABB

Data Scientist

ABB • Fulltime

Jul 2023 - Dec 2024 • 1 yr 5 mos

Data Scientist in the Finance Management Department. - Built and solely owned a financial budgeting data product from prototype to production: ingestion, cleaning, validation and UI (Python, SQL, Streamlit). - Cut budget goal calculation from 2+ months to minutes with a Python/Streamlit web app that replaced a manual Excel process. - Enabled self-service budget reports for Finance, Strategy, Risk, Management and C-level stakeholders. - Consolidated multi-year data across 10+ sectors (retail, mortgage, corporate, SME, deposits, treasury) into one financial view. - Cut cashback calculation from 1-2 days to under a minute by automating data collection and analysis in Python. - Analyzed 20,000+ customers in Oracle SQL, categorizing customer trends and transaction volumes. - Built forecasting and time-series models, with model selection and validation.