01
Evaluation Pipelines
Infrastructure that measures model behaviour so reliability is shown, not assumed.
Currently: KYC smart screening evaluation for Citi
Building AI systems production can actually rely on.
>
AI Engineer focused on evaluation pipelines, prompt engineering and reporting tooling that make model behaviour visible. Six years of backend engineering underneath.
LLM Evaluation
Prompt Engineering
Model Reporting
Async Event Pipelines
Clinical Order Ingestion
Production Integrations
LLM Evaluation
Prompt Engineering
Model Reporting
Async Event Pipelines
Clinical Order Ingestion
Production Integrations
LLM Evaluation
Prompt Engineering
Model Reporting
Async Event Pipelines
Clinical Order Ingestion
Production Integrations
LLM Evaluation
Prompt Engineering
Model Reporting
Async Event Pipelines
Clinical Order Ingestion
Production Integrations
[ ABOUT ]
[ HOW I BUILD ]
Evaluation first. Then everything else.
- Evaluation pipelines that measure how models actually behave.
- Prompt engineering for consistency and accuracy.
- Reporting tooling that makes model behaviour visible rather than a black box.
[ HOW I WORK ]
Plan early. Communicate clearly. No noise.
- Six years of backend engineering across healthcare, loyalty SaaS and food retail.
- Clinical order ingestion, async event pipelines and integrations that hold up under real production load.
- I plan early, communicate clearly, and don't create noise when things get hard.
6
YEARS OF BACKEND ENGINEERING
3
INDUSTRIES: HEALTHCARE, LOYALTY SAAS, FOOD RETAIL
LLM
EVALUATION AND PROMPT ENGINEERING
Citi
KYC SMART SCREENING EVALUATION
[ CAPABILITIES ]
[ From evaluation infrastructure to backend systems that hold up in production. ]
03
Reporting Tooling
Reporting that makes model behaviour visible rather than a black box.
[ CURRENT FOCUS ]
Currently building KYC smart screening evaluation infrastructure for Citi, closing prompt gaps that were driving inconsistent attribute extraction.
[ WORK ]
[ My experiences and my learnings across production AI and backend engineering. ]
- (01)
KYC Smart Screening Evaluation
AI ENGINEERING · CITI · CURRENT
Evaluation infrastructure for KYC smart screening. Identified and closed the prompt gaps that were driving inconsistent attribute extraction.
LLM EvaluationPrompt Engineering
- (02)
Model Behaviour Reporting
AI ENGINEERING
Reporting tooling that makes model behaviour visible to the team instead of a black box.
ReportingEvaluation
- (03)
Clinical Order Ingestion
BACKEND ENGINEERING · HEALTHCARE
Backend systems ingesting clinical orders in a healthcare environment.
BackendHealthcare
- (04)
Async Event Pipelines
BACKEND ENGINEERING
Asynchronous, event-driven pipelines running in production.
Event PipelinesBackend
- (05)
Production Integrations
BACKEND ENGINEERING · HEALTHCARE · LOYALTY SAAS · FOOD RETAIL
Integrations built to hold up under real production load.
IntegrationsProduction
[ ENGINEERING APPROACH ]
01
Plan Early
Understand the problem and its failure modes before writing code.
02
Measure First
Build evaluation first so behaviour is measured, not guessed.
03
Close the Gaps
Trace inconsistent output to its cause and fix it at the source.
04
Make It Visible
Report behaviour clearly so everyone sees what the system is doing.
05
Communicate
Clear updates and no noise, especially when things get hard.
AI ENGINEER
[ MEET BAHAUDDIN ]
AI ENGINEER · LLM EVALUATION · BACKEND SYSTEMS
AI Engineer building systems that turn LLMs into something production can actually rely on: evaluation pipelines, prompt engineering for consistency and accuracy, and reporting tooling that makes model behaviour visible rather than a black box.
Currently building KYC smart screening evaluation infrastructure for Citi, closing prompt gaps that were driving inconsistent attribute extraction.
That work sits on six years of backend engineering across healthcare, loyalty SaaS, and food retail: clinical order ingestion, async event pipelines, integrations that hold up under real production load.
I plan early, communicate clearly, and don't create noise when things get hard.
NOWAI Engineering, KYC Evaluation for Citi
6 YEARSBackend Engineering
HEALTHCAREClinical Order Ingestion
LOYALTY SAASBackend Engineering
FOOD RETAILBackend Engineering
[ LEARNINGS ]
[ Lessons from building AI and backend systems that real teams depend on. ]
Consistency starts with the prompt.
Inconsistent attribute extraction traced back to prompt gaps. Closing them is engineering work, not guesswork.
Visible beats black box.
Evaluation and reporting turn model behaviour into something a team can see and trust.
Production load tells the truth.
Six years of backend work taught one rule: integrations must hold up under real load.
Reliable
[ CONTACT ]
Hiring for AI reliability, LLM evaluation or backend systems? Send a message and I will reply personally.