I Find the AI Failures That Look Correct at First.
AI Evaluation and Quality Assurance specialist combining 15+ years of software QA, database, and data-validation experience with structured LLM evaluation, hallucination detection, instruction-adherence testing, and RAG quality auditing.
Where Rigorous QA
Meets AI Evaluation
“I evaluate complex information, validate system outputs, identify inconsistencies, and produce structured, evidence-based quality assessments.”
My AI evaluation work is built on 15+ years of experience in software QA, data validation, analytics, databases, and enterprise systems. Currently at HealthCare Digital Solution, Inc. (HCDS), I engineer production platforms (including BioProgressive.org and CS-Dashboard.hcds.ai), audit large-scale AI datasets (1,600+ Dolphin records), and lead data analytics and customer operations for Annie Insights.
Having end-to-end full-stack and customer operations experience provides essential engineering grounding: I evaluate AI outputs not in an abstract vacuum, but with an intimate understanding of how models integrate into live database schemas, user interfaces, and business workflows.
- Fikri Irawan Abdurahman · West Java, Indonesia
- fikri.irawan@gmail.com
- +62 813-2201-7333
- Timezone: UTC+7 (WIB) · Flexible overlap with US & EU working hours
- Availability: Remote contracts, freelance evaluation & project-based advisory
- Execution: Strict rubric adherence, fine-grained guideline compliance, asynchronous speed
“Fikri brings rare full-stack software and database discipline to AI evaluation. He systematically uncovers edge-case hallucinations and instruction non-compliance that automated evaluation pipelines consistently overlook.”
- Indonesian — Native
- English — Professional Working Proficiency
- Japanese — Basic
What I Do
Structured evaluation capabilities bridging enterprise quality assurance, data validation, and AI systems.
LLM Evaluation
Evaluate large language model responses against transparent rubrics to ensure outputs meet rigorous real-world standards.
AI Quality Assurance
Assess AI-assisted workflows using structured software QA methodologies, rigorous edge-case testing, and defect logging.
Data Quality
Validate enterprise datasets, inspect schema constraints, and verify operational anomalies across complex relational repositories.
Analytical Assessment
Transform complex telemetry and disparate outputs into structured findings, actionable evaluator feedback, and decision intelligence.
Built on 15+ Years of Quality Engineering
My AI evaluation work is built on more than a decade of experience in software validation, testing, data quality, database systems, analytics, and enterprise environments.
Telecommunications
Infrastructure monitoring, technical logs verification, and carrier network validation.
Software Engineering
.NET systems development, business rule verification, and output specification compliance.
Testing & QA
Comprehensive test case creation, defect lifecycle management, and QA automation.
Database & Data
Relational database administration, data anomaly detection, and schema validation.
Business Intelligence
Healthcare data platform validation supporting 100,000+ users with high reporting fidelity.
AI Evaluation
LLM response evaluation, hallucination detection, prompt refinement, and AI QA.
Selected Work & Live Projects
Production web platforms, large-scale AI dataset validation, and structured LLM evaluation case studies.
HCDS Customer Success Command Center
Full-stack operations and customer success platform built with Node.js 22, Express.js API, and Azure Static Web Apps. Integrates dual-design theme systems (Wave & Platform), Azure DB Query Gateway, ticketing analytics, escalation tracking, and system health telemetry for the ANNIE Orthodontic Platform and iSmartOffice systems.
BioProgressive International Web Portal
Official global web portal for the Foundation for Modern Bioprogressive Orthodontics (FMBO) engineered with semantic HTML5/CSS3, Vanilla JavaScript (ES6+), Node.js Azure Functions API, and Azure Static Web Apps. Features Google Identity Services (SSO), FMBODS distance study modules, Stripe donation processing, and an Annie VIP case tracking interface.
1,600+ Records Dolphin AI Dataset Quality Audit
Comprehensive quality audit and rubric-based evaluation across 1,600+ multi-turn reasoning and instruction-following pairs in the Dolphin fine-tuning corpus. Identified an 11.8% defect rate (189 of 1,600 audited records containing ≥1 qualifying defect: subtle reasoning drift, constraint omissions, or premise sycophancy), maintaining >96% inter-annotator calibration agreement against standardized gold rubrics.
Annie Insights — Analytics & Customer Operations
Dual-track operational role executing data analytics and high-tier customer service for the Annie Insights clinical orthodontic AI SaaS platform. Analyzed customer usage telemetry, monitored platform health and user adoption, triaged critical clinical escalations to engineering, and surfaced data-backed UX recommendations.
LLM Response Evaluation
Multi-response evaluation assessing two candidate AI outputs against a 6-point rubric (factual accuracy, relevance, completeness, constraint adherence, logic, and clarity).
Hallucination Detection & Fact Verification
Systematic claim decomposition and verification auditing an AI scientific summary, isolating unsupported assertions, false premise sycophancy, and numerical drift.
Prompt Evaluation & Refinement
Before-and-after demonstration revealing how architectural prompt constraints (role framing, schema definitions, and factuality parameters) eliminate vague outputs.
Healthcare Data Quality & AI Validation
Clinical discharge summary audit cross-checking AI text against raw EHR telemetry, isolating a 1,000x dosage unit error (50 mcg vs. 50 mg) and high-severity clinical contraindication omissions.
RAG Pipeline Evaluation & Faithfulness
Benchmarking vector retrieval precision, passage-level attribution, and answer groundedness across an enterprise technical infrastructure knowledge base.
How I Evaluate AI
A structured, six-step approach to assessing AI output quality and adherence.
UNDERSTAND
Deconstruct the prompt intent, user persona, explicit constraints, and expected output schema before examining model responses.
VERIFY
Cross-check discrete factual assertions against authoritative primary sources, verified databases, and established ground truth.
ANALYZE
Assess logical consistency, internal cohesion, completeness of key concepts, relevance, and cognitive load calibration for the target audience.
DETECT
Isolate subtle hallucinations, unsupported extrapolations, sycophantic premise acceptance, and unhedged overconfidence.
SCORE
Apply structured, multi-dimensional scoring rubrics (1–5 scale) to deliver objective, transparent, and reproducible quality metrics.
REPORT
Provide concise, evidence-based evaluator feedback, defect classifications, and actionable model tuning recommendations.
Professional Experience
Work across data engineering, platform development, AI quality evaluation, product analytics, and customer operations, supporting production systems serving 100,000+ platform users:
- AI & Dataset Quality: Evaluated 1,600+ multi-turn AI training records for instruction adherence, reasoning consistency, factual grounding, refusal quality, and recurring failure patterns using structured review criteria.
- Data Quality & Database Operations: Validate relational data, investigate operational anomalies, analyze query behavior, and maintain secure database systems and data dictionaries.
- Product Analytics: Analyze clinic telemetry, user retention cohorts, diagnostic session metrics, and customer workflows for Annie Insights, translating findings into actionable recommendations.
- Customer & Technical Operations: Triage high-priority clinical and technical issues, reproduce defects, document findings, and coordinate escalation to engineering.
- Web Platform Engineering: Developed and maintain production web platforms, including BioProgressive.org and the HCDS CS Command Center, using Node.js 22, Express APIs, Azure Static Web Apps, and database gateways.
Translated functional requirements into application logic and validation rules for .NET business applications. Tested application outputs against acceptance criteria, investigated defects, and maintained validation documentation supporting release quality.
Analyzed large datasets and transformed complex information into business insights and management dashboards. Evaluated data quality, reporting accuracy, and operational anomalies for decision support.
Performed structured QA for web and mobile applications, from test planning through defect documentation and regression validation. Created test cases from functional requirements and supported automation-oriented improvements to reduce repetitive manual QA cycles.
Developed custom business applications, relational database schemas, and data validation routines for regional enterprise clients. Executed end-to-end testing against user specification documents.
Monitored, evaluated, and tested telecommunications systems and operational network services. Reviewed operational reports, investigated service issues, and prepared technical documentation.
Skills & Tooling
- LLM Evaluation & Rating
- 1,600+ Dolphin Dataset Validation
- Fact Checking & Verification
- Hallucination Detection
- Instruction Adherence & Refusal QA
- Rubric-Based Quality Scoring
- Prompt Refinement & Architecture
- Quality Assurance (QA)
- Data Validation & Integrity Audits
- Root Cause Analysis (RCA)
- Test Case & Plan Design
- Business Rule Validation
- Process Compliance Verification
- Defect Investigation & Triage
- Node.js 22 & Express.js APIs
- Azure Static Web Apps (SWA)
- Vanilla JavaScript (ES6+)
- HTML5 & CSS3 Design Systems
- Google SSO (GIS) Integration
- Stripe Donation API Integration
- RESTful Services & Serverless HTTP
- SQL Server / PostgreSQL / MySQL
- Azure DB Query Gateway
- Product & Session Telemetry
- Power BI / Tableau / Grafana
- Microsoft Excel (Advanced)
- JSON Schemas & Data Pipelines
- Data Integrity Auditing
- Annie Insights CS Specialist
- Customer Health & Retention Tracking
- Clinical & Technical Escalation Triage
- SLA & Ticket Resolution Workflows
- Direct User Support Operations
- Cross-Functional Engineering Handoff
- Customer Experience Optimization
- ChatGPT (OpenAI)
- Claude (Anthropic)
- Google Gemini
- Microsoft Copilot
- Git / GitHub CI/CD
- Lighthouse Audit & Optimization
Complete Professional Profile
Download the resume most relevant to your opportunity.
AI Evaluation Resume
Tailored for AI labs, evaluation platforms, and research teams requiring structured LLM response rating, rubric scoring, prompt refinement, and hallucination detection.
Professional Resume
Comprehensive 15+ year record detailing enterprise database administration, software validation, healthcare analytics platforms supporting 100,000+ users, and business intelligence.
Frequently Asked Questions
Concise, structured answers on AI evaluation methodology, technical background, and project engagement.
Who is Fikri Irawan Abdurahman?
Fikri Irawan Abdurahman is an AI Evaluation Specialist, Data Analyst, and QA Professional with over 15 years of technical experience in software QA, enterprise database systems, and data validation across healthcare, orthodontic technology, and telecommunications. At HCDS, he builds production platforms (bioprogressive.org, cs-dashboard.hcds.ai), audits large AI datasets (1,600+ Dolphin items), and conducts analytics and customer service for Annie Insights.
AI Evaluation Specialist · QA & Data AnalystWhat does an AI Evaluation Specialist do?
An AI Evaluation Specialist systematically evaluates large language model (LLM) responses against structured criteria. Responsibilities include evaluating factual accuracy, measuring instruction adherence, detecting hallucinations, testing boundary constraints, scoring reasoning quality, and designing prompt architectures.
Methodology · LLM BenchmarkingHow does 15+ years of software QA and database experience apply to AI evaluation?
Enterprise software QA and database engineering provide the rigorous foundations for AI evaluation: atomic claim decomposition, boundary testing, root-cause defect analysis, and schema consistency audits. Having supported healthcare platforms serving 100,000+ users, Fikri evaluates AI outputs against rigorous empirical validation standards, ensuring ungrounded drifts are caught before production deployment.
QA Foundations · Empirical Output VerificationWhat live production projects and enterprise platforms has Fikri contributed to?
At HealthCare Digital Solution, Inc. (HCDS), Fikri engineered and maintains BioProgressive.org and the HCDS CS Command Center using Node.js 22 and Azure Static Web Apps, validated 1,600+ complex entries in the Dolphin AI dataset, and runs data analytics and customer support for the Annie Insights platform.
Live Enterprise Platforms · HCDSHow do you detect hallucinations and verify factual claims in AI outputs?
Hallucination detection utilizes atomic claim extraction: breaking complex AI output into discrete propositions, cross-checking each against authoritative primary sources (e.g. NASA, scientific journals, regulatory filings), and categorizing claims as Verified, Incorrect, Unsupported, or Hallucinated.
Fact Verification · Primary Ground TruthWhat AI evaluation case studies are demonstrated in this portfolio?
Alongside live production platforms, the portfolio features five independent demonstration case studies: (1) LLM Response Evaluation against a Six-Dimension rubric, (2) Hallucination Detection on James Webb Space Telescope data, (3) Prompt Evaluation for database migrations, (4) Healthcare AI Clinical Summary QA, and (5) RAG Pipeline Evaluation & Faithfulness Auditing.
5 Case Studies + Live PlatformsHow is AI evaluation different from traditional software QA?
Traditional software QA is deterministic: given input A, code executes predictable pathways to yield expected output B. Generative AI is probabilistic: given input A, a model produces varied responses B, C, or D that may all appear fluent yet differ in factual grounding, instruction adherence, or safety constraints. AI evaluation therefore requires probabilistic test suites, multi-dimensional rubrics, atomic claim extraction, false-premise stress testing, and automated assertion gates.
Deterministic vs. Probabilistic QAIs Fikri Irawan available for remote and freelance AI evaluation opportunities?
Yes, Fikri Irawan is available for remote, project-based, and contract opportunities worldwide in AI evaluation, AI trainer roles, LLM quality assurance, data quality audits, full-stack platform QA, and prompt engineering.
Global Remote · Contracts & Project Work
Let's Work
Together.
Available for remote and project-based opportunities in AI evaluation, AI quality assurance, data quality, and analytical work.