AI Analyst Agent (MCP)
Internal AI data analyst built with Gemini, Haystack, FastAPI and custom MCP servers. Teammates query Metabase, PostHog and BigQuery in plain language and build simple analyses and dashboards themselves.
$ ~ whoami
Hello, I'm
Danial_Azmi
Data Scientist & AI Engineer
Working across the full data stack, from data platforms to ML models to GenAI applications.
About —
About
I'm a data scientist working across the full data stack, building data platforms, training ML models and shipping GenAI applications in production. My focus is bridging the gap between advanced AI/ML systems and real business needs, making models useful, not just accurate.
Over the past ~4.5 years I've worked across fintech, media, consulting and edtech. I've migrated pipelines, stood up MLOps on the cloud and built recommendation engines. Lately I've been building RAG chatbots and AI agents that talk to company data through MCP.
edu: Bachelor of Computer Science (Data Science) — Multimedia University, Cyberjaya · 2017 – 2021
award: Yayasan Telekom Malaysia Scholar
fyp: Aspect-Based Sentiment Analysis on Steam game reviews
award: Yayasan Telekom Malaysia Scholar
fyp: Aspect-Based Sentiment Analysis on Steam game reviews
Experience —
Roles in detail
-
commit 7bd0490(HEAD -> main, origin/now)
Data Scientist @ Tupai AI
AI Teacher Sdn. Bhd.·Kuala Lumpur·
- Built the full analytics stack serving Marketing, Product and other teams, from data collection through to analysis, using PostHog, GA4, BigQuery, dbt, Prefect and Metabase on a medallion architecture.
- Developed a Bayesian recommendation system that creates customised learning plans for students.
- Created an AI agent with custom MCP servers (Gemini, Haystack, FastAPI) that works as an internal data analyst. It talks to Metabase, PostHog and BigQuery, so teammates can build simple analyses and dashboards themselves.
-
commit 76b16ba
Data Scientist @ REV Media Group
Kuala Lumpur·
- Developed an end-to-end RAG question-answering chatbot (Python, FastAPI, Haystack, GCP), improving customer engagement across multiple channels.
- Introduced CI/CD tooling and best practices, significantly streamlining the development lifecycle.
- Migrated on-premise ML models to GCP for better monitoring and observability.
- Developed a geo-located hyper-personalisation recommendation engine that drove a substantial increase in organic user engagement across products.
-
commit 43c532a
Data Science Consultant @ DataMicron Systems
DataMicron Systems Sdn. Bhd.·Kuala Lumpur·
- Led research and POCs to identify and validate AI/ML business use cases.
- Worked directly with clients to understand their challenges and propose AI/ML solutions to complex problems.
- Oversaw end-to-end project lifecycles, from model development and deployment to post-launch maintenance.
-
commit f9a160c
Data Scientist @ MoneyLion Malaysia
MoneyLion Malaysia Sdn. Bhd.·Kuala Lumpur·
- Migrated the team's Airflow and dbt ETL pipelines from AWS to Snowflake as part of a cloud database provider change.
- Designed, ran and monitored A/B tests to improve the NLU model and reduce operational costs.
- Analysed unstructured text data to surface topics and pain points and quantify them, giving a clearer picture of consumer behaviour.
- Built interactive dashboards that helped stakeholders make data-driven strategic decisions.
-
commit 0b0d9e2(tag: v0.1-intern)
Web Developer & Software QA Tester (Intern) @ Funai Malaysia R&D
Funai Malaysia R&D Sdn. Bhd.·Johor Bahru·
- Maintained and extended an internal web-based management and testing system built on Django.
- Worked with the QA team testing software for Android TVs.
Projects —
RAG Question-Answering Chatbot
End-to-end production chatbot using retrieval-augmented generation for a major Malaysian media group, serving customer queries across digital channels on GCP.
Adaptive Learning Recommender
Personalised learning-plan generator for an edtech platform. A Bayesian model adapts content order to each student's performance and history.
Geo-Localised Recommendation Engine
Hyper-personalised recommendations combining geo-location signals with behavioural data to lift organic article engagement.
More projects
-
Medallion Analytics Stack
Built the analytics stack for an edtech platform end to end, from event collection to dashboards, organised into bronze/silver/gold layers.
PostHog · GA4 · BigQuery · dbt · Prefect · Metabase -
On-Prem → GCP MLOps Migration
Moved legacy ML inference services to containerised deployments on Cloud Run, monitored via Vertex AI.
GCP · Docker · Kubernetes · Kubeflow · Vertex AI · CI/CD -
NLU A/B Testing Framework
Designed and monitored experiments on NLU model variants; reduced inference cost and improved intent accuracy.
Python · NLU · A/B Testing · Airflow -
ETL Migration (Redshift → Snowflake)
Migrated Airflow-orchestrated ETL and dbt models from AWS Redshift to Snowflake with zero downtime and full data validation.
Airflow · dbt · Snowflake · AWS -
Aspect-Based Sentiment Analysis (FYP)
NLP on millions of Steam game reviews, covering validation, preprocessing, neural networks and classical ML.
NLP · TensorFlow · Scikit-learn · Pandas -
AI/ML Consulting POCs
Client-facing POCs across computer vision, NLP and forecasting, covering scoping, building, hand-off and support.
Python · ML · FastAPI · Docker -
Internal QA Management System
Django web app supporting QA workflows for Android TV software.
Django · Python · JavaScript -
danbyte.my
This portfolio, built with FastAPI, Jinja2 and plain CSS/JS.
github.com/danialazmie/danbyte.my
Skills —
- languages
- PythonSQLJavaScriptBash
- ai-ml
- LLMsRAGGenAINLPMLOpsHaystackLangChainHugging FacePyTorchTensorFlowScikit-learnPineconeChromaDB
- data-eng
- dbtBigQueryAirflowPrefectFivetranSnowflakeRedshiftPandasMetabasePostHogGA4
- cloud
- GCPAWSAzure
- infra-tools
- FastAPIDockerKubernetesKubeflowNginxGitCI/CD
Contact —
Open to interesting problems. The inbox is the fastest route.
- danial.azmiee@gmail.com
- github
- github.com/danialazmie
- linkedin.com/in/danial-azmi
Try a command: help, about, projects, contact.