Skip to content
Sahil Swain

AI • Automation • Backend Engineering

I build reliable software systems for real-world workflows.

Senior backend engineer with 4+ years of experience across Java, Spring Boot, Microservices, AWS and Kafka, now applying that engineering discipline to AI-powered automation and agentic systems.

Explore the workStart a project

Lead Engineer experience • Enterprise systems • AI-native engineering

AI Agent Control Plane deterministic prototype

Control plane status: Standby.
agent-control-plane
Standby
Deterministic personal prototype · scripted states · no live model, policy engine, or tool integration

Plan · task #4471

  1. Parse objective into operations
  2. Resolve tools & constraints
  3. Propose operation op-4471
  4. Verify against policy
  5. Request human decision
  6. Execute under supervision

Verification

Policy check
—
Scope check
—
Risk level
—

Human decision checkpoint

op-4471 requests: write report to workspace

Audit trail

  • op-4468 · read-only query · executed · verified
  • op-4469 · write outside scope · rejected by policy
  • op-4470 · approved by operator · executed

Deterministic personal prototype — no live model or tool integration.

Software engineering experience
4+ years
HCL Technologies — supported by employment documentation
Lead Engineer
Targeted unit + integration coverage (raised from ~40% on owned services)
80%+
Documented project work: CitiBank, APL Logistics, The Vanguard Group
3 enterprises

Featured work

Proof over promises

Four pieces of evidence: a flagship personal system, two labeled automation prototypes, and a documented professional case study.

  • AI Systems

    Design AI workflows around explicit states, tools, verification, and human decision points.

  • Business Automation

    Turn repetitive manual processes into simple, inspectable workflows.

  • Backend Engineering

    Build APIs, distributed services, event-driven systems, data flows, and cloud-native infrastructure.

  • Engineering Quality

    Treat testing, failure handling, auditability, and maintainability as product features.

01Personal project

AI Agent Control Plane

Personal engineering project

A visual control surface for an agent framework that separates model reasoning from framework control: plan, verify, request human decisions where required, execute, and audit.

View case study
02Prototype — no client relationship claimed

Customer Follow-Up & Review Workflow

Business automation demo

A small workflow that helps a business follow up with customers after a completed service and makes it easier for customers to leave genuine feedback.

View case study
03Professional experience

Enterprise Backend Engineering

Professional case study

A multi-microservice financial file-ingestion platform — the Issuer Shared Service / CitiBank Data Lake — involving Java 17, Spring Boot, AWS S3, Apache Kafka, MySQL, Flyway, Docker and Helm.

View case study
04Prototype

Spreadsheet / Data Workflow Automation

Automation demo

Messy spreadsheet → deterministic validation and normalization → duplicate/anomaly report → clean output → optional lightweight reminder/dashboard.

View case study

Flagship · personal engineering project

AI Agent Control Plane

A visual control surface for an agent framework that separates model reasoning from framework control: plan, verify, request human decisions where required, execute, and audit.

“The model is not the framework. The framework controls the model.”
  • Execution state
  • Proposed operation
  • Verification state
  • Human decision checkpoint
  • Audit trail
  • Operation history
  • Risk / status information
Explore architecture
Modelreasoning onlyPlannerobjective → opsPolicy / Verifyscope · riskHuman checkpointapprove / rejectExecutorAudit trailproposalcandidate opverifiedapprovedrecordedrejectedThe model is not the framework. The framework controls the model.

Scroll horizontally to inspect the diagram at a readable size.

Business automation

Workflows that replace repetitive effort

Prototype demos running on deterministic synthetic data — built to be inspected, not just described.

Customer Follow-Up & Review Workflow

Prototype / Business Automation Demo
follow-up-workflow · demoPrototype · synthetic data

  • Alex Morgan

    Home cleaning · Service completed

    1. Service completed (next)
    2. Customer recorded (pending)
    3. Follow-up scheduled (pending)
    4. Neutral review request sent (pending)
    5. Event tracked (pending)

  • Priya Nair

    Lawn care · Service completed

    1. Service completed (next)
    2. Customer recorded (pending)
    3. Follow-up scheduled (pending)
    4. Neutral review request sent (pending)
    5. Event tracked (pending)

  • Daniel Osei

    Window washing · Service completed

    1. Service completed (next)
    2. Customer recorded (pending)
    3. Follow-up scheduled (pending)
    4. Neutral review request sent (pending)
    5. Event tracked (pending)

Event log

  1. No events yet — advance a step.

Policy: every customer receives the same neutral request. Review requests are never gated or suppressed based on sentiment.

Spreadsheet / Data Workflow Automation

Prototype

Messy spreadsheet → deterministic validation and normalization → duplicate/anomaly report → clean output → optional lightweight reminder/dashboard.

  • Deterministic validation rules
  • Normalization of inconsistent formats
  • Duplicate and anomaly reporting
  • Clean, inspectable output
Open the data-workflow demo

Enterprise engineering · professional case study

Issuer Shared Service — CitiBank Data Lake

Documented professional work, described at a public, non-confidential level.

  • Secure file-ingestion workflow
  • S3 access verification
  • Metadata / pipeline tracking
  • Service orchestration
  • Retry and exception handling
  • Flyway migrations
  • JaCoCo test coverage raised from ~40% to 80%+
  • Deployment collaboration with Helm / ECS / Autosys

Publicly described only at the engineering level supported by the résumé. No confidential implementation details, internal hostnames, credentials, customer-sensitive data, or proprietary business rules.

Read the case study

Technology domain

  • Java 17
  • Spring Boot
  • Microservices
  • AWS S3
  • Apache Kafka
  • MySQL
  • Flyway
  • Docker
  • Helm
  • Autosys

Also documented on the résumé

  • Tidal Watch — logistics platform, APL Logistics
  • Stox — investment & portfolio platform, The Vanguard Group

Method

AI-native engineering workflow

AI models are used as engineering collaborators for tasks such as design exploration, implementation, code review, testing assistance, refactoring, and documentation.

  • Planning / design exploration
  • Implementation
  • Code review
  • Testing assistance
  • Refactoring
  • Documentation

Verified work on this site

  1. 1Specification
  2. 2Kimi K3 implementation
  3. 3Read-only Command Code senior audit
  4. 4Targeted remediation
  5. 5Validation

Completed stages only. Unverified model use, final polish, and production deployment are not claimed.

Professional experience

Professional experience supported by employment documentation

  1. May 2025 – Sep 2025

    HCL Technologies

    • Lead Engineer

      May 2025 – Sep 2025

      Source: HCL experience letter + assignment letter

  2. Feb 2021 – Mar 2025

    Cadential Technologies

    • Software Engineer

      Documented Mar 2024 → Mar 2025

      Source: Cadential appraisal (7 Mar 2024) + resignation acceptance (relieved 10 Mar 2025)

    • Associate Software Engineer

      From 25 Jul 2022

      Promoted to Software Engineer at a date that is not established by the available documents.

      Source: Cadential offer letter (joining date 25 Jul 2022)

    • Software Intern

      15 Feb 2021 – 22 Jul 2022

      Source: Cadential internship letter

View professional verificationLinkedIn(link pending)GitHub(link pending)

Project details and metrics are supported by résumé / project records.

About

I'm a software engineer focused on backend systems, automation, AI-assisted workflows, and practical tooling.

My background is in enterprise software engineering with Java, Spring Boot, Microservices, AWS and Kafka. I'm now extending that foundation into AI systems that are structured, testable, and designed around explicit workflows rather than hype.

Technical depth

The toolbox, grouped by the problems it solves

Languages & Frameworks

  • Java 8 / 11 / 17
  • Spring Boot
  • Spring Security
  • Spring Data JPA
  • REST APIs

Data & Messaging

  • Apache Kafka
  • MySQL
  • Snowflake
  • Flyway

Cloud & Infrastructure

  • AWS S3 / EC2 / Lambda / CodePipeline / ECS
  • Docker
  • Kubernetes
  • Helm

Delivery & Quality

  • JUnit 5 / Mockito / Integration Testing
  • Jenkins / GitHub Actions / AWS CodePipeline
  • Git / Bitbucket / Jira / Agile / Scrum / Code Review

Python & Web

  • Python
  • Django

Have a workflow that should not require repetitive human effort?

Tell me what is manual today, what the output needs to look like, and where the workflow currently breaks.

Start a project

swainsahil079@gmail.com