AI Systems Engineer · Software Engineer · Agentic AI

Jeyanth V.

I build intelligent systems that work beyond the demo.

I engineer production-oriented AI systems, agentic workflows, and digital products—with a focus on reliability, practical problem-solving, and measurable impact.

Abstract network graphicDecorative connected nodes. It does not depict a real system.ABSTRACT / DECORATIVE
Chennai, IndiaSoftware Engineer · Rocket IndiaPython · MCP · React · Next.jsScroll

01Strengths

Where I do my best work

Five areas, each backed by work further down this page.

  1. 01

    Production AI & agentic systems

    Part of a team building multi-agent incident investigation, with reliability built in.

    Rocket India see the related work
  2. 02

    Multi-agent orchestration & MCP

    Team work on routing and consolidation around specialist agents that use MCP integrations.

    Rocket India see the related work
  3. 03

    LLM evaluation & observability

    Run traces and human-authored reviews used to catch hallucinations and gaps in AI-generated analysis.

    Rocket India see the related work
  4. 04

    Front-end & full-stack engineering

    React and Next.js CMS interfaces at Rocket India, Azure DevOps CI/CD work, and a full-stack product (React, Supabase, Cloudflare Workers) built solo.

    Rocket India · Kaasu see the related work
  5. 05

    Consulting & product development

    Leading freelance developers to deliver order, stock and distribution-tracking systems for SMEs; a personal-finance product built and deployed solo.

    JaffnaBusiness.com · Kaasu see the related work

02Featured work

Selected work, shipped and measured

Five professional contributions at Rocket India, then two independent products. Professional work is described at a high level; only measured results appear as numbers.

Professional engineering · Rocket India

Part of Rocket Companies. Described at a high level only: no internal screenshots, metrics, names or diagrams.

01ProfessionalTeam effortRocket India · Professional engineering

SRE Agent

Modular multi-agent orchestration

Turned incident triage into a coordinated team of specialist agents.

An AI-powered SRE assistant that helps investigate high-priority production incidents and works toward a root cause.

Contribution
Team effort: contributed to refactoring the existing SRE Agent into a modular multi-agent system for automated root-cause analysis.
Approach
  • Specialized agents, each responsible for one part of the investigation.
  • Stages that route the work and consolidate the findings.
  • Integrations that let agents retrieve operational data.
  • Reliability work so agents cope with transient failures.
Impact
  • As a team, helped responders reach a first root-cause hypothesis faster.
  • Python
  • MCP
  • LLMs
  • Multi-agent systems

Team effort

Demonstrates

  • System design
  • Agent orchestration
  • Failure handling
  • Iterating on traces

Internal work, described at a high level only.

02ProfessionalIndividual workRocket India · Professional engineering

Agent Evaluation & Observability Framework

Checking AI investigations against human reviews

Treats AI output as something to audit, not trust: human-written reviews set the bar.

An evaluation workflow that checks AI-generated investigations against human-authored reviews.

Contribution
Individual contribution: built the workflow that judges AI-generated investigations against human-authored reviews.
Approach
  • Compared agent output with human-written post-incident reviews.
  • Used execution traces to see how each investigation actually ran.
  • Looked for hallucinations and gaps, and fed findings back into prompts.
Demonstrates
LLM evaluation · Observability · Hallucination detection · Prompt optimization
Impact
  • Surfaced gaps in agent investigations and guided prompt improvements to make AI-assisted investigations more reliable.
  • LLM evaluation
  • Trace analysis
  • Prompt engineering

Internal work, described at a high level only.

03ProfessionalTeam effortRocket India · Professional engineering

Slack Context Ingestion Pipeline

Keeping agent context current

An agent is only as good as its context. This keeps that context fresh without anyone asking.

A scheduled pipeline that keeps an agent's operational context current from relevant team discussions.

Contribution
Team effort: contributed to automating the retrieval of relevant operational discussions and the refresh of agent context.
Approach
  • Scheduled retrieval of relevant operational discussions.
  • Refreshed agent context on a schedule.
  • Connected the current knowledge to AI-assisted reasoning.
Demonstrates
Data pipelines · Scheduled processing · API integration · Context management
Impact
  • Investigation agents work from current operational context.
  • Python
  • API integration
  • Data ingestion
  • Scheduled processing

Internal work, described at a high level only.

04CMS PlatformFront-end admin modulesProfessionalTeam effortRocket India

A content-management platform with administrative and user-management workflows.

Contribution
Team effort, front end only: built React and Next.js modules. No backend work was involved.
Approach
  • React and Next.js modules.
  • Front-end for role-based access control, CRUD workflows, dynamic search and user management.
Demonstrates
React and Next.js · Access-control interfaces · Search and CRUD UX · Working in a team
Impact
  • Contributed the interface for the platform's access-control, search and user-management capabilities.
  • React
  • Next.js
  • RBAC
05Mortgage Document IntelligenceUnstructured documents to automation-ready JSONProfessionalTeam effortRocket India

Prompt-driven extraction that turns unstructured mortgage documents into structured data for automation.

Contribution
Team effort: contributed prompt engineering for structured extraction, and to Azure DevOps CI/CD pipeline optimization with cross-functional teams.
Approach
  • Prompts that extract structured mortgage-domain data from unstructured documents.
  • Output converted into machine-readable JSON workflows.
  • Azure DevOps pipeline work focused on release processes and deployment reliability.
Demonstrates
Prompt engineering · Structured extraction · Domain modelling · CI/CD reliability
Impact
  • Extracted information made usable by downstream automation.
  • LLMs
  • Python
  • Prompt engineering
  • Document intelligence
  • Structured JSON extraction
  • Azure DevOps
  • CI/CD

Independent projects

Built and shipped outside my day job. Live links below.

Kaasu home screen on desktop in demo mode: sidebar navigation, a net-worth card showing sample data with XIRR, and quick actions for spending, income, investing and plans.
Kaasu home screen on a phone in demo mode: net-worth card with sample data, invested amount and gain, and quick actions for spending, income, investing and plans.

06IndependentIndividual workIndependent project · Solo product engineer

Kaasu

Independent personal finance product

End-to-end product ownership: domain modelling, local-first sync, security and deployment, built by one engineer.

A personal-finance app for Indian salaried investors: salary, spending, loans, goals and every investment in one place, with honest numbers derived from dated transactions.

Contribution
Designed and built the whole product alone: product definition, interface, domain logic, local-first sync and deployment.
Approach
  • Local-first: IndexedDB is the source of truth, so the app works offline; sign-in adds a per-user cloud copy.
  • Every figure is derived from dated transactions with FIFO lots and XIRR, stored as integer paise.
  • Supabase auth with row-level security, plus a Cloudflare Worker price API with JWT checks and rate limits.
Demonstrates
Product thinking · Domain logic and testing · Security by design · Shipping to real users
Impact
  • Live as a multi-user service with sign-up, a no-account demo and onboarding.
  • Domain logic covered by unit tests, and the database migration tested against real Postgres.
  • React
  • TypeScript
  • Vite
  • Tailwind CSS
  • IndexedDB (Dexie)
  • Supabase Auth & Postgres
  • Cloudflare Workers
  • Recharts
  • Vitest
  • PWA

Screens use the app's labelled sample data.

What the app does

What the deployed app does

  • Money: spending and income by category, category budgets, safe-to-spend per day, a spend-pace chart and bank-statement CSV import with auto-categorisation and duplicate skipping.
  • Bank accounts: balances side by side, a statement with running balance per account, transfers that never count as spending, and a 'match bank balance' correction.
  • Plans & debt: every SIP, RD, gold chit and EMI with instalments done and left, a 'what frees up when' timeline, a debt-free date and a prepayment simulator.
  • Portfolio: stocks, ETFs, mutual funds, deposits, bonds and gold; holdings, average cost, P/L and XIRR derived from transactions with FIFO lots.
  • India tax lots: per-lot holding period, unrealised STCG/LTCG, remaining LTCG exemption and a FIFO-aware zero-tax harvest helper. Rates and thresholds are editable.
  • Goals and savings wins: short and long-term goals with the monthly amount needed, plus milestones and streaks that respect reduced motion.
  • Command palette (Ctrl/⌘ K), encrypted JSON backup and restore, CSV export, light and dark themes, a privacy mode that blurs amounts, installable as a PWA.
Architecture and trade-offs

Architecture

Local-first. The browser's IndexedDB (through Dexie) is the source of truth, so the app works fully offline. Signing in adds a per-user cloud copy.

  • UI: React, TypeScript and Vite with Tailwind CSS, Recharts and a hand-built component set.
  • Domain layer: pure, unit-tested TypeScript for FIFO lots, XIRR, tax set-off, allocation, cash flow, payday and debt schedules.
  • Sync: table hooks write changes to an outbox; the engine pushes, then pulls by a server-assigned cursor. Last write wins and deletes are tombstones. Sample data never syncs.
  • Supabase: email magic-link and Google sign-in, one generic records table with row-level security forced so each user sees only their rows, a per-user write quota and self-serve account deletion.
  • Cloudflare Worker price API: CORS allow-list, per-IP and per-user rate limits, Supabase JWT verification, AMFI NAVs and Google Finance quotes with a Yahoo Finance fallback.
  • Static frontend on Vercel with a strict Content-Security-Policy, HSTS and frame protections.

Data handling

  • Money is stored as integer paise; dates are plain calendar dates. No floating-point rupees.
  • Holdings and balances are never stored as truth: they are derived from dated transactions, so every figure can be traced back.
  • A price is never shown without its date and source. Manual prices always work.
  • Backups can be encrypted in the browser (AES-256-GCM with a PBKDF2-derived key) and are validated before a restore replaces anything.

Engineering trade-offs

  • Last-write-wins sync instead of field-level merging: simpler to reason about for a single person on a few devices, at the cost of occasionally losing a concurrent edit.
  • Google Finance has no public API, so quotes come from its quote page with Yahoo as a fallback. Either can change, so the UI always says when a holding can't be priced.
  • Tax modelling is explicit about scope: surcharge, carried-forward losses and 2018 grandfathering are listed as not modelled rather than approximated.
  • Kaasu is a personal tracker, not investment or tax advice, and the app says so.
JaffnaBusiness.com home page on desktop: headline 'Your partner in business growth' beside the Arivu elephant mascot in front of a Jaffna coastline, with registered-company and 12-services facts below.
JaffnaBusiness.com services page on a phone: 'Twelve services, grouped by where you are' with chips for Start, Manage, Grow and Digital & learning.

07IndependentIndividual workIndependent consulting · Live redesign

JaffnaBusiness.com

Consulting and product development

Started from the client's problem: twelve services reorganized around the customer's journey, in two languages.

A redesign of the public website for Jaffna Business (Pvt) Ltd: twelve services organised around the client's journey, an English/Tamil interface, and Arivu, an assistant that answers first questions and hands off to the team.

Contribution
As freelance software consultant: information architecture, interface design and implementation of the public redesign, alongside wider consulting on accounting automation and digital workflows.
Approach
  • Twelve services regrouped into a four-stage client journey, each with its own page and a comparison view.
  • A full English / தமிழ் interface with hreflang alternates for search engines.
  • Arivu, an assistant with suggested questions that hands visitors to a person or WhatsApp.
Demonstrates
Consulting · Information architecture · Bilingual UX · Delivery coordination
Impact
  • Redesign live on a public Vercel deployment; the production domain still serves the earlier site.
  • Structured data, per-page metadata and responsive layouts from phone to desktop.
  • Vite
  • Single-page app
  • Bilingual content (EN / தமிழ்)
  • JSON-LD structured data
  • Responsive design
  • Vercel

A public preview: jaffnabusiness.com itself still serves the earlier site. Client figures and testimonials are not claimed as my results.

How the redesign works

Information architecture

Instead of a flat list, the twelve services are grouped by where the visitor is in their business journey.

  • 01 Start: business registration, foreign company registration, trust and charity formation.
  • 02 Manage: company secretarial, tax and legal support, bookkeeping.
  • 03 Grow: consulting, market research, import and export, HR consulting.
  • 04 Digital & learning: website creation and stock-market courses.
  • Each service has its own page, and a comparison view puts all twelve side by side.

English and Tamil

A language switch in the header swaps the whole interface between English and Tamil, and the page declares its English and Tamil alternates with hreflang so search engines can serve the right one.

Enquiry paths

  • A 'Book a Consultation' call to action in the header and through the page, leading to a dedicated booking page.
  • Click-to-call and WhatsApp links for visitors who would rather talk.
  • A Ctrl K search in the header.

Arivu, the assistant

Arivu, the firm's elephant mascot, fronts JaffnaAssistant: a chat panel with suggested questions ('Which service fits my business?') and free-text input, available in both languages, which passes visitors to a person or WhatsApp when they need one.

Status and scope

Status

The redesign is live on its Vercel deployment. The production domain jaffnabusiness.com still serves the earlier site, so the redesign is presented here as a public preview rather than the production site.

03For hiring teams

If you are hiring, start here

Short version for a hiring manager. The detail is in Featured work.

I help teams get AI systems from prototype to something they can trust: agentic workflows, evaluation against human ground truth, and the interfaces and delivery around them.

  • Measure before claiming

    I only state results I can verify, and I keep private details private.

  • Design for failure

    Retry and recovery are part of the design, not an afterthought.

  • Check the machine against people

    Human-written reviews are the ground truth for model output.

04Experience

Where I have been building

  1. January 2025 to Present

    Chennai, India

    Rocket India, part of Rocket Companies

    Software Engineer

    Part of a team building production AI systems and the platforms around them: agentic incident investigation, MCP integrations, context ingestion, front-end CMS engineering, mortgage document intelligence and CI/CD. I owned the evaluation and observability framework individually. The detail lives in Featured work.

    • Agentic incident investigation for production incidents
    • Evaluation and observability of AI-generated investigations
    • Operational context ingestion for agent workflows
    • Front-end CMS modules in React and Next.js
    • Prompt engineering for structured document extraction
    • Azure DevOps CI/CD pipeline optimization with cross-functional teams

    This is internal work. Descriptions are high level only, with no diagrams, metrics or internal names, and there are no links to internal tools, repositories, dashboards or data.

    Company links are public context only and do not show internal work.

  2. June 2023 to Present

    Hybrid / Chennai, India

    HackerKID & GUVI Geek Networks, IITM Research Park

    Freelance Mentor / Junior Mentor

    Mentoring students in Python, problem-solving, Generative AI, Data Science and full-stack development. Most of the work is technical communication: finding the explanation that makes a concept click, then helping the student build it.

    • Helped students understand programming and technical concepts.
    • Supported problem-solving and practical implementation.
    • Used interactive learning content and responsive doubt resolution to support student engagement.
  3. May 2025 to Present

    Remote

    JaffnaBusiness.com

    Freelance Software Consultant

    Consulting small and medium-sized businesses on accounting automation and digital transformation, and leading a team of freelance developers to deliver business systems: an end-to-end order and stock management system for star-impex.de, tracking tools for chicken distribution and reselling businesses, and similar systems for other SMEs.

    • Translated business requirements into practical software solutions.
    • Supported accounting automation and digital workflow improvements.
    • Led 5 client projects with a rotating team of freelance developers, including an end-to-end order and stock management system for star-impex.de.
    • Delivered distribution and reselling tracking systems for chicken-trade SMEs, and similar business systems for other small companies.
    • Delivered dashboards and web applications across these engagements.
    • Created AI-assisted business development videos using modern AI tools.
    • Contributed to the evolution of JaffnaBusiness.com as a digital business-services platform.

05Technical skills

The toolkit

Grouped by what I use them for, not ranked.

01 Languages

  • Python
  • TypeScript
  • JavaScript
  • Java
  • SQL

02 Frameworks

  • React
  • Next.js
  • Tailwind CSS
  • FastAPI

03 AI & agent systems

  • MCP servers
  • Agent SDK
  • LLM evaluation
  • Prompt engineering
  • Document intelligence
  • NLP
  • LLM observability

04 Cloud & DevOps

  • Azure DevOps
  • Git
  • CI/CD
  • YAML
  • Cloudflare Workers
  • Vercel

05 Databases

  • PostgreSQL
  • MySQL
  • Supabase
  • IndexedDB

06Education

Foundations

  • 8.45CGPA

    Rajalakshmi Engineering College

    B.Tech. in Artificial Intelligence and Machine Learning

    2021 – 2025
    Chennai, India

  • 95%HSE score

    Vels Vidyashram

    HSE, Computer Science and Mathematics

    2020 – 2021
    Chennai, India

07Contact

Have a meaningful engineering problem? Let's build something useful.

ramanjeyanth3@gmail.com