Why Aung
Technical Product Lead for HydraX AI Transformation
Built AI-with-controls patterns in production and delivered under real regulatory constraints.
View Evidence MatrixExecutive Summary
Aung is a product and engineering leader with 13+ years building AI-native products and regulated fintech systems end-to-end. Co-founded an AI FinOps platform shipping LangGraph multi-agent orchestration, policy-aware RAG, risk-tiered HITL approval workflows, and guardrails to live pilots in 6 weeks. Before that, he led product delivery across a central-bank-regulated digital lending platform serving 300K+ users with audit logging, regulatory reporting, and 5 payment gateway integrations. He brings the exact intersection HydraX needs: deep AI workflow architecture fluency, practical compliance delivery experience, and rigorous outcome-tied KPI design.
Core Argument
Three pillar reasons this candidacy fits HydraX's transformation requirements.
Built required AI workflow patterns in production
- ›LangGraph multi-agent orchestration with risk-tiered HITL routing (Shadow → Supervised → Autonomous)
- ›Policy-aware RAG with regulatory document curation, citation quality controls, and relevance scoring
- ›Confidence-weighted signal extraction — LLM for nuanced signals, deterministic heuristics for control events
4 live SME pilots shipped in 6 weeks with 173+ eval tests
Delivered regulated products with compliance as first-class constraint
- ›CBM-regulated digital lending platform — mandatory reporting, blacklist management, data privacy controls
- ›Immutable audit logging (officer, action, full payload, timestamp) with dual-write sync to core banking
- ›RBAC with 12 granular permissions across 150+ API endpoints with OAuth/JWT security
300K+ users, 5 loan products, 32% incident reduction via RCA loops
Executes fast with measurable outcomes and control discipline
- ›Discovery to live pilots in 6 weeks with outcome-tied KPI instrumentation
- ›21% loan application completion improvement via funnel diagnosis and UX redesign
- ›44% feature delivery lead time reduction through architecture and process upgrades
4-layer eval framework tracking cycle-time, error rates, and handoff load
Evidence by Job Requirement
Every requirement from the TPL job description mapped to concrete proof with confidence levels.
| Job Requirement | Details | Aung's Evidence | Source | Confidence |
|---|---|---|---|---|
| AI-assisted workflow design (LLM + RAG + guardrails + HITL) | Design policy-aware workflow intelligence with risk-tiered decisioning and failure-mode controls | Designed LangGraph multi-agent architecture with risk-tiered HITL routing (Shadow → Supervised → Autonomous), approval gate criteria at each transition, and deterministic rule checks protecting critical workflow control boundaries | ARiA (FingentiX) | High |
| Policy-aware retrieval with citation quality and freshness controls | Operationalize policy retrieval with evidence-backed, explainable recommendations | Implemented RAG with regulatory document curation, category filtering, citation quality controls, and explainability — recommendations grounded in authoritative, traceable sources | ARiA (FingentiX) | High |
| Event-level audit logging and evidence design | Event-level logging of context, suggestion, decision, approver, and outcome for audit-ready evidence packs | Designed immutable event logging (officer, action, full payload, timestamp) and dual-write data patterns deployed in production fintech for compliance-ready audit evidence | AMAY Management | High |
| Deterministic rules vs model-based decisioning tradeoffs | Ensure deterministic rule checks protect critical control boundaries while leveraging model intelligence where appropriate | Designed confidence-weighted signal extraction — LLM for nuanced behavioral signals, deterministic heuristics for critical control events — with registry-based policy rules extensible without code changes | ARiA (FingentiX) | High |
| Roadmap ownership from problem framing to production rollout (90/180/365) | Translate strategy into executable delivery plans with phased milestones, gates, and dependency maps | Defined phased milestones, gate criteria, and dependency maps — delivered AI pilot from discovery to 4 live SME pilots in 6 weeks with outcome-tied KPI instrumentation | ARiA (FingentiX) + AMAY | High |
| Cross-functional delivery in compliance-intensive environments | Coordinate Product, Engineering, Compliance, Ops, Legal, and leadership — resolve tradeoffs quickly and transparently | Coordinated product, engineering, compliance, ops, and executive stakeholders across regulated fintech delivery and 30+ enterprise client engagements | AMAY + MM-Digital-Solutions | High |
| KPI frameworks tied to execution quality and risk | Track KPIs beyond output volume — cycle-time, exception load, rework, adherence, audit quality | Built eval frameworks tracking cycle-time, error rates, and human handoff load; tied roadmap investments to measurable risk reduction | ARiA (FingentiX) | High |
| API-driven architectures and integration patterns | Reason about API-driven architectures, data contracts, orchestration, and observability | Managed 150+ API endpoints with OAuth/JWT security; led 8-microservice GCP deployment (Terraform IaC, Cloud Run + GCE) with structured release and incident governance | AMAY Management | High |
| Risk-tiered decisioning (inform / justify / approve / block) | Implement risk-tiered decisioning with clear approval gates and failure-mode controls | Designed risk-tiered HITL routing (Shadow → Supervised → Autonomous) with approval gate criteria at each transition and deterministic rule checks at critical boundaries | ARiA (FingentiX) | High |
| MAS/CMS or equivalent regulatory context exposure | Operating familiarity with Singapore financial regulatory frameworks | No direct MAS/CMS operating history. Closest equivalent: CBM (Central Bank of Myanmar) regulated digital lending — mandatory reporting, blacklist management, policy-gated approvals | AMAY Management | Gap |
| Tokenization / custody / market infrastructure familiarity | Domain knowledge of digital capital-markets infrastructure — issuance, custody, dealing/exchange | No direct tokenization/custody/dealing shipping history. Transfer basis: regulated financial product lifecycle management, multi-stakeholder approval workflows, audit-ready system design | — | Gap |
Design policy-aware workflow intelligence with risk-tiered decisioning and failure-mode controls
Designed LangGraph multi-agent architecture with risk-tiered HITL routing (Shadow → Supervised → Autonomous), approval gate criteria at each transition, and deterministic rule checks protecting critical workflow control boundaries
Operationalize policy retrieval with evidence-backed, explainable recommendations
Implemented RAG with regulatory document curation, category filtering, citation quality controls, and explainability — recommendations grounded in authoritative, traceable sources
Event-level logging of context, suggestion, decision, approver, and outcome for audit-ready evidence packs
Designed immutable event logging (officer, action, full payload, timestamp) and dual-write data patterns deployed in production fintech for compliance-ready audit evidence
Ensure deterministic rule checks protect critical control boundaries while leveraging model intelligence where appropriate
Designed confidence-weighted signal extraction — LLM for nuanced behavioral signals, deterministic heuristics for critical control events — with registry-based policy rules extensible without code changes
Translate strategy into executable delivery plans with phased milestones, gates, and dependency maps
Defined phased milestones, gate criteria, and dependency maps — delivered AI pilot from discovery to 4 live SME pilots in 6 weeks with outcome-tied KPI instrumentation
Coordinate Product, Engineering, Compliance, Ops, Legal, and leadership — resolve tradeoffs quickly and transparently
Coordinated product, engineering, compliance, ops, and executive stakeholders across regulated fintech delivery and 30+ enterprise client engagements
Track KPIs beyond output volume — cycle-time, exception load, rework, adherence, audit quality
Built eval frameworks tracking cycle-time, error rates, and human handoff load; tied roadmap investments to measurable risk reduction
Reason about API-driven architectures, data contracts, orchestration, and observability
Managed 150+ API endpoints with OAuth/JWT security; led 8-microservice GCP deployment (Terraform IaC, Cloud Run + GCE) with structured release and incident governance
Implement risk-tiered decisioning with clear approval gates and failure-mode controls
Designed risk-tiered HITL routing (Shadow → Supervised → Autonomous) with approval gate criteria at each transition and deterministic rule checks at critical boundaries
Operating familiarity with Singapore financial regulatory frameworks
No direct MAS/CMS operating history. Closest equivalent: CBM (Central Bank of Myanmar) regulated digital lending — mandatory reporting, blacklist management, policy-gated approvals
Domain knowledge of digital capital-markets infrastructure — issuance, custody, dealing/exchange
No direct tokenization/custody/dealing shipping history. Transfer basis: regulated financial product lifecycle management, multi-stakeholder approval workflows, audit-ready system design
18-Month Roadmap Fit
How Aung's track record aligns with each transformation phase's success targets for the Job.
AI-Fluent Foundation
- ›25–40% improvement in requirement-to-spec and delivery prep cycles
- ›Team-wide AI operating playbooks in product/engineering/ops/compliance
- ›Policy-grounded retrieval and documentation support with measurable quality
- ›Governance baseline: clear AI usage boundaries, approvals, and traceability
AI-Assisted Core Ops
- ›Production pilot in at least one high-volume workflow
- ›Deterministic policy checks integrated into critical transitions
- ›Risk-tiered human approval routing implemented
- ›End-to-end audit evidence packets generated from workflow events
AI-Native Workflows
- ›Expanded AI-assisted workflows across issuance/custody/trading/post-trade
- ›Meaningful reduction in manual exception handling and rework
- ›No critical control breakdowns attributable to AI design
- ›Leadership-ready control and performance dashboards
AI-Fluent Foundation
- ›25–40% improvement in requirement-to-spec and delivery prep cycles
- ›Team-wide AI operating playbooks in product/engineering/ops/compliance
- ›Policy-grounded retrieval and documentation support with measurable quality
- ›Governance baseline: clear AI usage boundaries, approvals, and traceability
AI-Assisted Core Ops
- ›Production pilot in at least one high-volume workflow
- ›Deterministic policy checks integrated into critical transitions
- ›Risk-tiered human approval routing implemented
- ›End-to-end audit evidence packets generated from workflow events
AI-Native Workflows
- ›Expanded AI-assisted workflows across issuance/custody/trading/post-trade
- ›Meaningful reduction in manual exception handling and rework
- ›No critical control breakdowns attributable to AI design
- ›Leadership-ready control and performance dashboards
Technical Depth
Architecture credibility: what HydraX needs, what Aung built, and why the patterns transfer.
API & Integration Architecture
API-driven architecture reasoning, data contracts, orchestration, and observability across tokenisation/custody/dealing
150+ API endpoints (GraphQL + REST), OAuth/JWT security, 8-microservice GCP deployment (Terraform IaC, Cloud Run + GCE), 5 payment gateway integrations
Data Models & Event Logging
Event-level logging of context, suggestion, decision, approver, outcome for audit-ready evidence packs
Immutable audit logging (officer, action, full payload, timestamp), dual-write sync to core banking, RBAC with 12 granular permissions
Rule/LLM Decision Boundary
Clear boundaries between deterministic policy rules and model-based decision support across critical transitions
Confidence-weighted signal extraction — LLM for nuanced behavioral signals, deterministic heuristics for critical control events, registry-based policy rules extensible without code changes
Guardrails & Failure-Mode Testing
Structured AI workflow boundaries, approval gates, and failure-mode controls with measurable confidence
6-mode failure testing protocol (hallucination, prompt injection, legal boundary violations) with severity ratings and retest cycles. 173+ eval tests as production KPIs.
Regulated Delivery Proof
Three case cards demonstrating delivery in regulated, AI-workflow, and enterprise contexts.
AMAY Management
CBM-regulated digital lending platform — 300K+ users, 5 loan products, 5 payment gateways, credit scoring, regulatory compliance
Central bank mandatory reporting, blacklist management, data privacy controls, cross-functional delivery under active regulatory oversight
GraphQL collection API with immutable audit logging and RBAC (12 permissions), 150+ endpoints with OAuth/JWT, 8-microservice GCP deployment with Terraform IaC and CI/CD
Shipped compliance-grade platform at scale with measurable delivery improvements and incident reduction
ARiA by FingentiX
AI-native FinOps platform automating Order-to-Cash workflows for SME finance teams using AI agents
Regulated financial workflows requiring explainability, audit trails, and human approval at critical control boundaries
LangGraph multi-agent architecture with risk-tiered HITL routing, policy-aware RAG pipeline, confidence-weighted signal extraction, 4-layer eval framework (173+ tests)
Shipped from discovery to 4 live SME pilots in 6 weeks with full AI governance stack
MM-Digital-Solutions
Product/engineering org delivering enterprise-grade projects across 15+ industries
Multi-stakeholder coordination across state telcos, listed conglomerates, and MNCs with cross-border delivery requirements
40+ enterprise projects for 30+ organizations; B2B regulated payments platform (3 bank gateways); omnichannel commerce SaaS (0→PMF); scaled org to 35+ people
Demonstrated execution governance at enterprise scale with seed funding, investor QBRs, and strategic pivots
Gap & Bridge
Transparent acknowledgment of domain gaps and the concrete plan to bridge them.
The Gap
No direct tokenisation/custody/dealing shipping history
No production experience with digital asset issuance, custody infrastructure, or dealing/exchange systems
No MAS/CMS direct operating history
Regulatory experience is CBM (Central Bank of Myanmar) — structurally similar compliance patterns but not Singapore-specific frameworks
The Bridge Plan
Structured domain ramp-up with compliance pairing
Weeks 1–6: paired working sessions with HydraX compliance and operations leads to absorb tokenisation lifecycle, custody requirements, and MAS/CMS specifics
First 6–8 weeks domain immersion outputs
Deliverables: workflow pain map, policy retrieval baseline, and pilot lane specification — demonstrating domain absorption through shipped artifacts, not just study
Transfer of proven architecture patterns into capital-markets context
The same patterns (immutable audit logging, policy-gated approvals, risk-tiered HITL, compliance-embedded delivery) apply directly — the domain vocabulary changes, the architecture does not
First 90 Days at HydraX
Execution plan mapped directly to the Job's expected deliverables.
Workflow & Control Discovery Baseline
- Baseline workflow pain map across issuance/custody/dealing/post-trade
- Prioritized pilot lane selection based on business value + control risk
- Stakeholder map and alignment cadence established
AI Operating Playbooks & Approval Matrix
- Role-specific AI operating playbooks (product, engineering, ops, compliance)
- Approval matrix: where AI assists vs. where human approval is mandatory
- Initial policy retrieval + evidence standards definition
Pilot Lane Spec & Instrumentation
- Pilot spec for one high-volume workflow with KPI and control metrics
- Gate criteria instrumentation setup for Phase 2 decision
- Deterministic rule checks and approval routing for pilot scope
Pilot Launch & Governance Cadence
- Pilot live in production with monitoring and control dashboards
- Governance cadence and reporting template operational
- First structured review cycle with measured outcomes
Workflow & Control Discovery Baseline
- Baseline workflow pain map across issuance/custody/dealing/post-trade
- Prioritized pilot lane selection based on business value + control risk
- Stakeholder map and alignment cadence established
AI Operating Playbooks & Approval Matrix
- Role-specific AI operating playbooks (product, engineering, ops, compliance)
- Approval matrix: where AI assists vs. where human approval is mandatory
- Initial policy retrieval + evidence standards definition
Pilot Lane Spec & Instrumentation
- Pilot spec for one high-volume workflow with KPI and control metrics
- Gate criteria instrumentation setup for Phase 2 decision
- Deterministic rule checks and approval routing for pilot scope
Pilot Launch & Governance Cadence
- Pilot live in production with monitoring and control dashboards
- Governance cadence and reporting template operational
- First structured review cycle with measured outcomes
Decision Synthesis
This role requires a rare intersection: product ownership, AI architecture fluency, and compliance-embedded delivery discipline.
Aung's pattern history maps to 9 of 11 job requirements at High confidence, with 2 domain gaps transparently addressed through a concrete bridge plan. The transferable patterns — risk-tiered HITL, policy-aware retrieval, immutable audit logging, compliance-gated workflows — are structurally identical to what HydraX needs to move from AI-Fluent to AI-Native.