Six years of enterprise platforms across fintech, logistics and AI. Backend engineering, cloud on AWS, Azure and GCP, and multi-agent systems on LangGraph and MCP, running in production, not in demos.
Right now, systems I architected serve 100K+ requests a day at 99.95% uptime, across 15+ delivered projects.
Manual operations cut60%
Deploy time, monolith to services−75%
Slowest Oracle queries800 → 120ms
Systems in production.
Platforms that move real money, real shipments and real traffic every day.
Inebura
Fintech AR platform · active
AI-powered accounts receivable: invoice processing, payment reconciliation and predictive AR. A real-time architecture that flags cash-flow risk before it happens, run by agentic workflows.
Java 21
Spring Boot
Python
LangGraph & MCP
Kafka
vLLM & RAG
AWS, cloud-portable
70%less manual work
95%reconciliation accuracy
$500K+saved a year
$50M+invoices processed
Inebura · Collections
INV-24081 · sample customer
72delay probability
Paid 9 days late on the last 4 invoices
Disputed 2 of the last 10
No reply to email in 6 days
Next action
OwnerCollections
ChannelCall AP lead
DeadlineFriday
Cash collected, last 12 weeks
Illustration with sample data.
Integrably
Logistics middleware · active
One API for every carrier: payload transformation, protocol bridging across REST, SOAP, SFTP, webhooks, CSV and EDI, smart routing and unified tracking.
1 APIfor all carriers
24/7observability
API Gateway
Infrastructure · active
Authentication, rate limiting, routing and monitoring in front of the microservices.
10K+peak requests/s
99.95%uptime
ComixFlip
Platform · completed
A digital comics platform with a custom reader, AI recommendations and offline reading as a PWA.
10K+readers
50K+comics
Task Orchestrator
Agentic AI · personal
An agentic system that reads a task in plain English, plans it and runs it on its own.
24/7autonomous
0babysitting
Inside Inebura, layer by layer.
A multi-tenant SaaS that predicts late payment before it happens, with hard data isolation for every enterprise client and a security posture aligned with ISO 27001 and SOC 2.
Portability
Integrations
Data
Application
Edge
Edge
Managed load balancing, TLS termination and WAF protection. Every request resolves its tenant before it touches the app.
The subdomain maps to the tenant’s own schema
The JWT carries a tenant claim, so a token can never cross tenants
Zero-downtime rolling deploys on AWS
Application
A React 19 app with three role-gated apps in one code-split build, and two Spring Boot services on Java 21 virtual threads.
Service one: receivables, invoicing, communications and BI
Service two: identity, ledger and tenancy
Containerised on AWS with autoscaling
Data
Managed MySQL with regional HA: a synchronous standby with automatic failover, plus a read replica.
One schema and one encryption key per tenant
Redis and Caffeine as a two-level cache
Kafka streams drive the real-time dashboards
Integrations
Everything a finance team’s documents and messages pass through, each bound to its tenant.
S3 signed URLs for documents
Exchange and SMTP set up per client
Scheduled SFTP document intake
WebSockets bound to token and tenant
Portability
Vendor-neutral by design, so the cloud is a deployment decision rather than a redesign.
GCP: Cloud Run or GKE, Cloud SQL, Memorystore, Pub/Sub, Document AI
Azure: an equivalent managed service for every layer
Take the database down.
Failover is a design, not a hope. Kill the primary and watch the synchronous standby take over while reads keep flowing.
A simulation of the designed behaviour. In production the drill is rehearsed, automated and written into the DR plan.
MySQL · regional HAAll healthy
db-a · zone 1takes writes
primary
db-b · zone 2synchronous copy
standby
db-r · zone 3serves reads
read replica
Press “Take the primary down” to start.
Decisions that mattered.
A schema per tenant, not a tenant column
Each client gets its own MySQL schema and key: isolation you can show an auditor, not just assert in a WHERE clause.
Kafka instead of polling
Event-driven engines compute dashboard metrics on the stream, so dashboards stay live without hammering the transactional database.
Failover as a design, not a hope
The standby is promoted automatically, the replica keeps reads flowing, and the recovered node rejoins.
Audit as architecture
Versioning across about 150 entities maps onto ISO 27001 Annex A and SOC 2 controls. The system produces its own compliance evidence.
The AI, in three layers.
Layer 1 · identity
Behaviour DNA
A live profile per customer: payment pattern, preferred channel, disputes and risk trend, computed deterministically.
Layer 2 · prediction
Delay prediction
A 0–100 delay score on every invoice, with its top three reasons and a named action, owner and deadline.
Layer 3 · orchestration
Collections playbook
An autonomous engine that plans who to contact, when and on which channel, built on the two layers below.
Six years, one through-line.
Building the systems that money and freight move through.
Balram SinghSolutions Architect · New Delhi
TanServ
2025 to now
Solutions Architect / Tech Manager
Microservices serving 100K+ daily requests at 99.95% uptime. Predictive AR and multi-agent workflows on LangGraph and MCP. Monolith to microservices.
−60% manual ops
−75% deploy time
Safexpress
2023
Senior Software Engineer
APIs handling 50K+ daily transactions, real-time notifications on WebSockets and Redis, and test coverage raised from 40% to 85%.
Oracle 800ms → 120ms
Mastercard
2023
Full Stack Developer, contract
Enterprise solutions for global payment processing: 12+ features across 3 sprints with zero production incidents.
−45% dashboard load
Techginia Global
2021
Java Developer / Scrum Master
8 API modules for 20K+ daily users at 99.7% uptime, and 2 developers mentored to promotion.
Deploys 2h → 15min
Concentrix
2020
Software Engineer
150+ API endpoints across several projects, with a 98% code-review approval rate and 3 major releases.
The full stack.
Cloud & infra
AWS
Azure
GCP
Kubernetes
Docker
Terraform
CI/CD
Backend
Java 21
Spring Boot
Python
Node.js
REST
GraphQL
Microservices
AI & agents
Multi-agent orchestration
MCP
LangGraph
LangChain
RAG
LLM evals
Vector DBs
Data & messaging
PostgreSQL
Oracle
Redis
Kafka
WebSockets
EDI
Architecture
System design
Security
Scalability
Event-driven
API design
Practices
Scrum
DevOps
Mentoring
Tech leadership
Roadmapping
Let’s build something that scales.
Architecture reviews, AI agents in production, or a platform from scratch. Write to me, or explore my freelance services.
Websites that make an impression. MVPs that get to market. AI that does useful work. Work directly with me, from the first architecture decision to launch, with a clear scope and milestone-based fees.
Custom adapters and field mappings have to be built again for each integration.
1 / 3
Conceptual product walkthrough. No live API calls.
02 / TRACE · SERVICE DESK PLATFORM
From a lost email to a promise kept.
Shared inboxes hide ownership. Manual triage eats the day. Deadlines slip quietly. I built Trace to bring tickets, automation, SLA policies and customer visibility into one service desk.
CUSTOMER PORTALRequest receivedIllustrative request · no live customer data
The request has a place to live.
An email becomes a tracked ticket. Replies stay on the same history instead of disappearing into a shared inbox.
1 / 4
A simplified walkthrough of Trace’s ticket lifecycle. No emails are sent or tickets created.
03 / THE GOOD HIRE · HIRING AUTOMATION
From scattered applications to a shortlist with evidence.
I built The Good Hire to bring the early rounds of hiring into one workflow: structured applications, role-specific work samples, reminders and evidence-backed scoring. Recruiters get the context behind the shortlist, and a person makes the final decision.
Indicative starting prices for a focused scope. INR rates are for India; USD rates are for international projects, not a live currency conversion. Final fees, payment milestones and any applicable taxes are confirmed in the proposal. Domain, hosting, paid APIs and ongoing support are quoted separately.
You bring
The idea, the vision and the domain expertise. You know the problem cold.
I bring
The complete technical build: architecture, engineering, cloud and AI, end to end.
A clear path to launch.
Brief
Tell me the idea and the problem it solves.
Scope
Agree on deliverables, fees and payment milestones.
Build
I own and ship the full tech stack, with you.
Grow
Launch, hand over and plan any ongoing support.
From idea to live product.
Capabilities we can scope into your project, from the first prototype to launch and ongoing improvements.
1/8Prototype and website
Prototype and website
A working prototype and a landing site, so the idea is validated before any heavy investment.
Build, test, refine
Every module goes round the loop until it is solid, secure and production-grade.
Tech and cloud selection
AWS, Azure or GCP, matched to your scale, compliance needs and budget. Deliberate choices, not hype.
Full product architecture
Services, data flow, APIs, security and scaling paths, mapped before production code is written.
AI where it pays
Agents, automation and LLM workflows exactly where they save real work.
Cost optimisation
Right-sizing, autoscaling and cutting waste across every service, so the product never overpays.
Delivery and launch
CI/CD, monitoring and observability in place for a clean go-live. Never a hand-off and vanish.
Support and managed service
Optional ongoing support, updates and iteration, scoped and priced separately.
Who it’s for.
Solo founders
You have the vision and the domain expertise, but no technical co-founder or team.
Early startups
Pre-seed teams who need serious engineering before they can raise or hire.
Entrepreneurs
People with a real problem worth solving and the commitment to build it right.
The questions you’re thinking.
How is the final price decided?
We discuss your requirements, then I send a written proposal covering deliverables, timeline, fees and payment milestones. The listed prices are starting points; integrations, complexity and scope determine the quote.
Do you ask for equity?
No. These are paid freelance engagements with agreed project fees. You keep your company ownership.
Who owns the code and IP?
Your project’s custom code and deliverables transfer to you on full payment, under our written agreement. Third-party tools and open-source components retain their own licences.
What if the scope changes?
We agree on the extra cost and timeline before adding work. Milestone reviews keep progress visible and give you a chance to course-correct.
Is hosting and support included?
Deployment setup is included where specified in your proposal. Domain, hosting, paid APIs and ongoing maintenance are separate costs. Support can be scoped as a follow-up engagement.
Let’s build your next thing.
It takes about two minutes. The more specific you are about the problem and where you are now, the better.
1I read every pitchYou hear back within a few days.
2A short discovery callWe check the vision and the fit.
3Terms in writingScope, fees and milestones agreed before I build.
