Senior AI Engineer · Generative AI · Agentic AI · LangGraph · RAG
Wahyu Fatur Rizky
I build production Generative AI: agentic workflows, RAG systems and LLM platforms that are evaluated, observable and safe to run in regulated environments.
Semarang, Central Java, Indonesia
- Open to Senior AI Engineer roles worldwide
- Available immediately
- Ready to relocate
- Requires visa sponsorship
- On-site, hybrid or remote
- years in software engineering
- 11+
- years in software engineering
- years shipping production GenAI
- 3+
- years shipping production GenAI
- bank branches served
- 120
- bank branches served
- countries reached by platforms I lead
- 60
- countries reached by platforms I lead
About
Production GenAI, engineered like any other critical service
I'm a Senior AI Engineer with more than 11 years building software, including about four years of remote contract work alongside my Computer Science degree, and over three years shipping Generative AI that runs in production rather than in demos. I design agentic workflows and Retrieval-Augmented Generation (RAG) with LangGraph and LangChain, across hosted models (OpenAI, Azure OpenAI, AWS Bedrock, Gemini, Claude) and open-weight models (Llama, Mistral) served with vLLM and Ollama.
My background is backend engineering in Python and Java, so I treat LLM features like any other production service: evaluated before release, traced in production, resilient to provider outages and deployed on Docker, Kubernetes and OpenShift.
- Banking GenAI at Bank Mandiri: a RAG assistant for 120 branches and agentic request triage with human approval, PII masking and audit logging.
- GenAI delivery at Accenture: an agentic AI operations assistant on Azure AI Foundry for a global telecom client across 60 countries, integrated with ServiceNow, Splunk and Dynatrace through MCP tools.
- Reinvention with Agentic AI, Accenture × Stanford HAI.
What I build
Agentic AI workflows
LangGraph state machines with tool calling, agent memory, multi-agent routing and human-in-the-loop approval for risky actions.
Production RAG
Hybrid retrieval (vectors + full-text), re-ranking, context management and cited answers that refuse instead of hallucinating.
Model-agnostic LLM platforms
Gateways that route between hosted and open-weight models by cost, latency and data residency, with automatic fallback.
LLM evaluation & observability
Golden datasets, regression gates in CI, and tracing with LangSmith, Langfuse and OpenTelemetry.
Security & governance
PII masking, guardrails, prompt-injection defence, access controls and audit logging for regulated industries.
Enterprise integration
Connecting LLMs to core systems, ITSM and observability tools (via MCP), databases and internal APIs through Python and Java services.
Experience
11+ years, from full-stack to banking-grade GenAI
Sep 2024 – Present
Semarang, Indonesia · Hybrid
Custom Software Engineering Associate Manager
Hands-on GenAI Technical Lead
Accenture
- Lead teams of up to 18 engineers delivering enterprise apps, cloud integrations and GenAI for telecom, oil & gas, energy, banking, FMCG and airline clients, with teams across Indonesia, India, the Philippines, the USA and the UK.
- Global telecom client (60 countries): led the production release of an agentic AI operations assistant built with LangGraph and Microsoft Agent Framework on Azure AI Foundry (Azure OpenAI, Python, LangChain).
- Connected ServiceNow (incl. MID Server), Splunk, Dynatrace/SCOM and Everbridge through reusable MCP server tools and an API abstraction layer, so agents triage alerts, open incidents and notify teams via Microsoft Teams and email.
- Built an operations monitoring dashboard and bridge orchestration for incident calls, then reused these patterns for other clients to cut operating cost and raise productivity.
- Own architecture through production for 5 live GenAI applications serving 2,000+ enterprise business users, with LangGraph agents (tool calling, memory, human-in-the-loop checkpoints) that automate 8+ business processes.
- Raised RAG answer accuracy by 25% with hybrid retrieval and re-ranking, and cut p95 latency by 35% and LLM token cost by 30% through caching, streaming and model routing, at 99.9% availability.
- Technical Manager, global energy & utilities: cloud automation, data integration, analytics and AI workflows, with up to 20% better operational efficiency and up to 30% faster time-to-market.
- Technical Lead, financial services (Philippines): technical delivery, critical defect resolution and unit-testing practices to meet coverage thresholds.
- Project Lead, global consumer goods: digital platforms for health, hygiene and nutrition brands across 60+ countries, with UK stakeholders. Technical Lead, global electronics: consumer electronics platforms, with South Korea stakeholders.
May 2024 – Sep 2024
South Jakarta, Indonesia · On-site
Full Stack Software Engineer
Kopra IT Digital Wholesale Channel Delivery Dept.
PT Bank Mandiri (Persero) Tbk.
- Delivered agentic request triage with human approval before any sensitive action, plus PII masking and audit logging for banking compliance, cutting manual handling time by 40%.
- Extended the RAG assistant to wholesale channel use cases on the Kopra platform.
- Integrated third-party and internal services within a 700+ developer organisation, and reduced reported defects by 30% and production downtime by 40%.
- Handled production troubleshooting and releases, and wrote documentation that shortened onboarding.
Nov 2023 – May 2024
South Jakarta, Indonesia · On-site
Full Stack Software Engineer
IT Digital Branch Channel Delivery Dept.
PT Bank Mandiri (Persero) Tbk.
- Built a RAG assistant for branch staff used by 1,500+ staff across 120 branches, with PII masking and audit logging.
- Self-hosted Llama and Mistral for data residency, with guardrails on model output.
- Built Angular front-ends and back-office tools for the New Branch Delivery System, and wireframes, user flows and mockups for Livin' back-office features.
- Followed Mandiri engineering and security standards with audit-ready documentation.
Oct 2022 – Nov 2023
Jakarta, Indonesia · Remote
Full Stack Lead Engineer
Axiata Digital Labs
- Led 18 engineers building web and mobile products for XL Axiata and Axiata Group; about 90% of the work served XL Axiata.
- Delivery lead for the XLSMART Employee Portal and Boost, Axiata's all-in-one fintech app.
- React, Next.js, Vue and Angular on the front end, Java/Spring Boot on the back end, working with Android and iOS teams.
- Raised team productivity by 15% and process efficiency by 20% (Jira delivery metrics).
- Shipped the team's first production LLM feature: customer-support ticket summarisation with output validation and rule-based fallback.
- Built a RAG knowledge assistant adopted by 50+ engineers that cut documentation lookup time by 30%.
Jan 2015 – Oct 2022
Singapore · Remote
Senior Software Engineer (React, React Native, Vue, Angular)
Mentor, Glints Expert Class · Contract
Glints
- Remote contract engineer placed with client teams; part of this period ran alongside my Computer Science degree (2015 – 2019).
- WhatsHalal (2020 – 21): Assurance Traceability Management System for halal certification. Erloom Digital Venture (2020): AutoTrust for Toyota Trust / AUTO2000. KOTA Digivice (2019 – 20): Muaman Insurtech.
- 2019: lead front-end for SUN Education (Digital Niaga Solusindo), an LMS for Marshall Cavendish Education, Cling (Meteor Inovasi Digital) and UI/UX prototypes for Virtualspirit Malaysia. 2018 – 19: humanitarian data visualisation for iMMAP.
- Also FWD Insurance, DocDoc, Yummy Corp, Ace Resource (Asian Agri, Apical) and Nabati Group.
- Stack: React, React Native, Vue, Angular, Flutter, Redux, Webpack, Jest, Mocha and Enzyme. Mentored engineers in Glints Expert Class.
Open-source projects
Production patterns, tested in CI
Each repository runs its full test suite and evaluation gates on every push, offline by default and against Gemini when a key is provided.
LLM Evaluation & Reliability Framework
Fails CI if the pass rate drops, a passing case breaks, or p95 latency or cost per case exceeds budget. Retries with backoff and jitter, circuit breakers, a fallback chain and rate limiting; JSONL tracing and OpenTelemetry; Gemini, Claude, OpenAI and Ollama, plus an offline fault-injecting mock.
- Python
- Gemini
- GitHub Actions
- OpenTelemetry
Agentic AI Workflow Starter
LangGraph banking support: an LLM router hands off to card, account, dispute and FAQ agents, with interrupt()-based human-in-the-loop approval, customer identity injected from graph state, least-privilege tools and SQLite checkpoints. Gemini by default.
- LangGraph
- LangChain
- FastAPI
- Docker
Production RAG Pipeline
HNSW vector and full-text search on pgvector fused with Reciprocal Rank Fusion, re-ranking, validated citations and safe refusal. The CI eval gate holds a 100% hit rate and 100% grounded answers.
- pgvector
- PostgreSQL
- LangChain
- FastAPI
Model-Agnostic LLM Gateway
OpenAI-compatible gateway with smart, fast and private route aliases (private allows only self-hosted models, enforced at startup), PII redaction (cards, emails, phones, NIK), SHA-256-hashed virtual keys with per-key rate limits and monthly USD budgets, caching, SSE streaming and Prometheus metrics. Tested in CI with the official OpenAI Python SDK.
- FastAPI
- vLLM
- Ollama
- Prometheus
Skills
Tools I use in production
- LLM frameworks
- LangGraph
- LangChain
- Microsoft Agent Framework
- LlamaIndex
- DSPy
- MCP
- Models
- OpenAI
- Azure OpenAI
- Azure AI Foundry
- AWS Bedrock
- Gemini
- Claude
- Llama
- Mistral
- vLLM
- Ollama
- GenAI techniques
- Agentic AI
- Multi-agent orchestration
- RAG
- Tool calling
- Prompt engineering
- Context management
- Structured outputs
- Languages & backend
- Python
- Java
- Spring Boot
- SQL
- TypeScript
- FastAPI
- Node.js
- REST APIs
- Microservices
- Data & retrieval
- pgvector
- PostgreSQL
- FAISS
- OpenSearch
- Pinecone
- Weaviate
- Milvus
- Observability & quality
- LangSmith
- Langfuse
- OpenTelemetry
- LLM evaluation
- Golden datasets
- E2E testing
- Platform
- Docker
- Kubernetes
- OpenShift
- CI/CD
- AWS
- Azure
Certifications
- McKinsey.org Forward Program, McKinsey.org (Jun 2026)
- Primerli Industry Foundations (Apr 2026): Banking, Insurance, Software, Semiconductors, Consumer Electronics, Medical Devices, Oil & Gas, Oil & Gas Upstream, Oil & Gas Midstream & Downstream
- Reinvention with Agentic AI, Accenture × Stanford HAI
- HackerRank (2024): REST API (Intermediate), SQL (Advanced), Python (Basic), Frontend Developer (React), Angular (Intermediate), Node.js (Basic)
- Machine Learning with Python, Ministry of Communication and Information Technology, Indonesia (2020)
Education
Bachelor's degree, Computer Science, Universitas Sumatera Utara (Aug 2015 – Aug 2019), GPA 3.58 / 4.00
Languages
- Bahasa Indonesia (Native)
- English (Professional Working)
- Javanese (Native)
- Arabic (Elementary)
- Bahasa Melayu (Limited Working)
Mentoring & Community
Teaching what I ship
Contributor
Program Asah led by Dicoding with Accenture Indonesia
Feb 2026
Recognition plaque, Bandung.
Advisor
Pijak in collaboration with IBM SkillsBuild (Dicoding)
3 capstone mentoring sessions, 4.67 / 5 participant satisfaction, 91% programme graduation rate.
Instructional Facilitator
Rakamin Academy × PT Bank Syariah Indonesia
Jan – Jun 2024
Speaker
Sinau Koding Webinar #15
May 2025
Essential IT skill sets for 2025.
Super Mentor
Dealls
Instructor
Glints Expert Class
React, React Native and JavaScript.
Honours & awards
- USU Employer Reputation 2024 (Oct 2024)
- 3rd place, Leadership Project Poster National Competition, Universitas Indonesia (2018)
- 1st place Best Stand Expo and 2nd place Business Plan, ASENET International Competition (2018)
- Top 5 Duta Mahasiswa USU (2017)
- 1st place Social Entrepreneurship, Sharia Economic Learning Forum XIV, Udayana (2017)
- 1st place National Scientific Essay, KOIMPLUS UNIMED (2017)