01 / 07 ·
I build systems that become products.
I'm Vipin Singh, a CTO, tech founder and AI and solution architect. I design configurable workflow platforms with AI inside human-approved boundaries, across insurance, HRMS, logistics and enterprise software.
16+Years in technology
7+Years in insurance tech
7Apps, one workflow core
02 // neural map
One mind. Thirteen technologies. Seven applications.
Experience feeds the stack, and the stack powers every application. Scroll to grow the network, hover a node to trace its path, open an application to see it working.
03 // projects universe
Seven applications. One core.
Every project runs on the same workflow core: state, rules, approvals and an audit trail you can trust. Open one to fly to it and see its screens, flow and stack.
03 // project briefs
The same seven projects, in plain words.
Read a brief here, or open any project in the universe to see its screens.
01 · Insurance technology · Ready to go live
Configurable insurance workflow platform
Proposals from the agent desk arrive here and run as one configurable flow. The platform covers workflow orchestration, dynamic forms, product configuration, masters, rules and decisioning, document processing, case management and integrations.
Read the brief
What it does
- Intake from the agent desk, then document evaluation and extraction.
- Rules and decisioning as versioned, configurable decision tables.
- Low-confidence fields go to a person; medical disclosures go to an underwriter.
- An append-only audit trail, so every decision can be replayed.
How it flows
IntakeDocumentsRulesRouteHuman reviewPolicy system
Where AI fits
Document AI extracts each field with a confidence score. AI confidence never decides the outcome on its own; it only decides who looks next. Below 0.85, a person checks the field.
Built with
02 · Insurance sales · Production
Agent policy desk
An agent-facing app for new business. The agent picks the customer, picks the product, configures the plan and submits the proposal straight into underwriting.
Read the brief
What it does
- Find the customer by name, mobile or PAN, with KYC status up front.
- Compare products, then configure sum assured, term and riders.
- The rating engine prices base premium, riders and GST as you go.
- Submitting sends the proposal straight into underwriting as a new case.
How it flows
CustomerProductPlanSubmitUnderwriting case
Checks before submit
KYC, product age limits and tele-medical requirements are checked before the proposal leaves the desk, and the rating engine prices base premium, riders and GST.
Built with
03 · Document management · Production
Document management system
Policies, claims, employee files, contracts and freight documents in one place, with versions, retention rules and access control, and every document linked to the case it belongs to.
Read the brief
What it does
- Policies, claims, employee files, contracts and freight documents in one place.
- Versions with full history, classification and retention rules.
- Access control, with an access log on every record.
- Every document is linked to the case it belongs to.
How it flows
Check inClassifyVersionApproveRetain and log
Rules at work
Classification, retention and access are rules on every record, and every view is logged. The policy schedule here is the same case the insurance workflow approved.
Built with
04 · Recruiting technology · Production
Modern ATS and resourcing platform
Jobs, candidates, recruiter assignment, interviews and decisions on the same workflow core, plus client management and reporting, with AI helping recruiters instead of replacing the record.
Read the brief
What it does
- Jobs, candidates and recruiter assignment in one pipeline.
- AI summaries and match notes on each candidate.
- Interviews, feedback and decisions tracked stage by stage.
- Client management, approvals and reporting built in.
How it flows
SourcedScreeningInterviewDecisionClient approval
Where AI fits
AI drafts candidate summaries and match notes to help recruiters. The recruiter edits them, and the record and the decision stay with people.
Built with
05 · AI operations · Production
AI + workflow automation
AI for extraction, reasoning and exceptions. Deterministic workflow for state, approvals, guardrails and auditability. A person sits at the gate between them.
Read the brief
What it does
- LLMs and agents take the ambiguous work: extraction, reasoning and exceptions.
- Deterministic workflow keeps state, approvals and guardrails explicit.
- A human approval step sits between AI output and anything that ships.
- Every step stays auditable, from model output to final approval.
How it flows
ExtractReasonExceptionsHuman approvalWorkflow stateAudit
The boundary
Use AI where ambiguity exists. Use deterministic software where correctness, state and compliance matter. Nothing ships past the human gate unchecked.
Built with
06 · Logistics technology · Ready to go live
Freight management
Bookings, shipments, carriers and milestones from port to door. Freight documents are read by AI, checked by rules and filed in the DMS.
Read the brief
What it does
- Bookings, shipments and carriers from port to door.
- Milestones and a live status for every shipment.
- Freight documents extracted by AI, with a confidence score on every field.
- Fields are checked by rules and filed in the DMS; unclear ones go to a person first.
How it flows
BookingShipmentDocument AIRules checkDMSDelivered
Where AI fits
High-confidence fields flow straight into the shipment. The smudged container number scores 0.62, so it becomes an exception for a person, not a guess in the system.
Built with
07 · HR technology · Production
Benefits system
One configurable HR system for benefits enrolment, employee loans, allowances and statutory leaves, with eligibility as rules and payroll as the final step.
Read the brief
What it does
- Benefits enrolment with eligibility written as rules.
- Employee loans with eligibility checked automatically.
- An approval chain from manager to HR, finance and payroll.
- Allowances and statutory leave balances in one place.
How it flows
RequestEligibility rulesManagerHRFinancePayroll
Rules at work
Service length, EMI-to-pay ratio and active loans are checked automatically. Approval runs manager, HR, then finance, and the EMI lands in payroll.
Built with
04 // live AI pipeline
AI where ambiguity exists. Deterministic software where it matters.
One insurance case, end to end, the way the platform actually routes it: AI extracts, rules decide, a person checks what the model was unsure about.
- IntakeProposal from the agent deskwaiting
- Document AIExtract fields with confidencewaiting
- Rules v14Decision table, first match winswaiting
- Human reviewOnly what the model doubtswaiting
- Policy systemApproved update, append-onlywaiting
Case APP-20417queued
audit.logappend-only
Three rules I build every system on.
- rule 01 in the case: Rules v14Configuration over hard-coding
Frequently changing business processes should become versioned configuration wherever practical.
- rule 02 in the case: Human reviewAI with boundaries
Agents can reason and act, but workflows, permissions, approvals and auditability remain explicit.
- rule 03 in the case: audit.logProduction over demos
Architecture includes security, observability, failure handling, deployment and operational ownership.
05 // journey
Technology became the medium.
The direction of my work has moved from implementing software toward designing systems, platforms and products.
git log --reversemain
- 16+ yearsTechnology to architecture
Enterprise engineering, architecture, integrations, delivery and technical strategy.
- 7+ yearsInsurance to deep domain systems
Deep insurance technology experience translated into reusable workflow and platform concepts.
- ProductsProjects to platforms
Product directions spanning insurance workflow, insurance sales, PIVV, document processing, ATS and HRMS/Benefits.
- HEAD → NowIndependent technology builder
Focused on products, consulting, technical leadership and experiments at the intersection of enterprise systems and AI.
stack.yaml6 layers
- intelligence
- LLMs, RAG, AI agents, Document AI
- interface
- React 19, single-page apps, nginx, iOS and Android
- process
- BPMN, DMN, rules engine, forms, RPA
- services
- Java 21, Spring Boot 3, OpenAPI, REST, microservices
- data
- MongoDB, Amazon DocumentDB, PostgreSQL
- infrastructure
- AWS S3, AWS SES, Docker, distributed tracing
# Process and decision models use open standards, so they stay yours.
Complex operations. Clear systems.
I turn operational complexity into software that can be configured, audited, scaled and eventually productized.
- Workflow & case systems
Queues, approvals, forms, rules, case management, audit trails and human-in-the-loop operations.
apps/ - Insurance platforms
Agent sales, application processing, document evaluation, underwriting workflows and policy-system integration.
apps/ - Benefits & HRMS
Benefits, employee loans, allowances and statutory leaves, with eligibility as rules and payroll integration.
apps/ - Logistics & documents
Freight management, shipment milestones, document management and AI extraction for document-heavy operations.
apps/
From the blog.
Four long-form briefings so far on AI architecture, AI with boundaries and decision automation, and nine more in the archive.
06 // still building
Still building. Still exploring.
Technology is the main thread, but curiosity doesn't stop at the stack.
Have a hard problem? Let's talk about the system.
I work as a CTO, tech founder and AI and solution architect: architecture reviews, AI workflow builds and technical due diligence. Tell me what you are working on, and I will tell you honestly whether I can help.