Premkumar ASay hello
CHENNAI, INDIA ↗

✳ Hello, I’m Premkumar

Full Stack
Engineer.

I build production AI agents and the full‑stack products around them: Python APIs, LangGraph workflows, and React interfaces. Founding engineer at AlphaTales.

Explore my work Resume
2+ years
At AlphaTales
50+
REST API endpoints
4+
Production AI workflows

Hi, I'm Premkumar. I'm a full stack engineer. I build AI-powered SaaS products.

MEET PREMKUMAR · PLAY INTRO
Privacy

The microphone starts only when you tap. Your browser may process voice online. Typed and spoken questions, with recent replies, go to the portfolio’s answer service on Cloudflare Workers AI, with OpenRouter as a backup. Questions are logged without personal details to improve the answers. Topic buttons play existing recordings.

SCROLL TO EXPLORE 01 — INTRODUCTION

02 — Selected work

What I’ve built.

A SaaS product, deployed agent workflows, developer tools, and a mobile prototype. Each case study explains the problem, my contribution, and what works today.

01
Built with the team

AlphaTales

From a product idea to a development plan.

My part: Co-built end to end; web interfaces, Python APIs, and AI workflowsOutcome: Review and export requirements, architecture plans, and tasks.50+ REST endpoints4+ production AI workflowsLive product
Project detailsRead case studyClose case study

Work at AlphaTales

Founding Engineer · Full Stack

AlphaTales helps teams research a product idea and turn it into requirements, architecture plans, and development tasks.

Explore productExplore the public workflow

My work on this project

I co-built AlphaTales end to end with the team. My work spans Python APIs, AI workflows, login, live progress updates, and React screens for reviewing and exporting plans.

How it works

  1. 01

    Idea

    Describe the product idea and the problem it should solve.

  2. 02

    Research

    Review the idea and gather research to inform the plan.

  3. 03

    Plan

    Generate requirements, architecture recommendations, and development tasks.

  4. 04

    Export

    Download the plans and implementation details for developers and coding assistants. AlphaTales calls this a Dev Pack.

Public product; team source code is private.

01

Engineering decision

Kept feature details in a shared format from AI generation through storage, review, and export, so the same information could move between those steps.

02

What works today

Users can review generated plans and export requirements, architecture details, and tasks for developers and coding assistants.

Keeping feature details consistent

Worked on how feature behavior, edge cases, permissions, and verification details pass between the AI output, API, review screen, and exported files.

Handling expired sessions and retries

Worked on refreshing expired OAuth tokens and handling retries in agent integrations. Request identifiers account for the input, so repeated calls can be distinguished from changed requests.

FastAPIPostgreSQLLangGraphAzure OpenAIAzure Service BusSocket.IOWorkOSNext.js 16MCP

Built with the AlphaTales team. This is a simplified workflow example; the company’s source code is private.

Explore productExplore the public workflow
02
Team agent workflows · VPS hosted

AlphaTales Agents

Specialist agents for product work, lead research, and service health.

My part: Agent profiles, task handoffs, Azure monitoring, CRM fixes, and deploymentOutcome: Hosted workflows for coding handoffs, CRM writes, and health alerts.3 specialist agentsHosted on a VPSHuman review built in
Project detailsRead case studyClose case study

Company project · Hosted on a VPS

Contributor · Agent workflows & deployment

Give product tasks, lead research, and service monitoring their own agent workflows, with clear handoffs and human review where needed.

My work on this project

I helped build and deploy these workflows with the AlphaTales team. My contributions include the Product Manager agent profile and task handoffs, the Azure Health agent, HubSpot lead-write fixes, and VPS runtime configuration.

Three specialist workflows

  1. Product Manager

    Routes product tasks to coding workers, reviews evidence, and coordinates approval and PR-repair handoffs.

    My contributionPublished the installable profile, added specialist subagents, and strengthened the task and PR-repair loop.

    Coding handoffs · Review decisions

  2. Lead Generation

    Researches target accounts with Apollo and writes qualified company information to HubSpot. Outreach has separate approval rules.

    My contributionFixed CRM writes to use properties supported by the HubSpot schema, with regression tests.

    Target accounts · CRM records

  3. Azure Health

    Checks service health, App Service state, and Service Bus. Sends confirmed problems and recoveries to Microsoft Teams.

    My contributionAdded the monitoring profile and improved read-only checks, scheduling, and the dedicated Teams alert channel.

    Health checks · Teams alerts
VPS runtime

Helped deploy the workflows on a VPS and improved runtime configuration, environment-secret loading, and scheduled health checks.

01

Why separate agents

Each profile has a specific job, its own configuration, and clear limits. The Product Manager agent coordinates coding work; lead generation handles account research; the health monitor runs service checks.

02

What works today

The workflows are hosted on a VPS. The repository includes installable profiles, deployment and verification scripts, and tests for agent contracts, CRM writes, and health monitoring.

PythonHermesMCPApolloHubSpotAzure CLILinuxMicrosoft Teams

Built with the AlphaTales team. Company source code and operational details are private.

03
Idea & co-building

AlphaOptimizer

Shorter tool outputs, with the original kept available.

My part: Proposed the idea; helped build, debug, and improve itOutcome: Retrieve the original output after receiving a shorter result.My idea30 focused tests passedPublic contribution
Project detailsRead case studyClose case study

AlphaTales tool · Public source

Idea originator · Co-builder

AlphaOptimizer helps Codex handle large command and tool outputs. It keeps useful lines visible and the original available for a limited time. Jev ranks relevant text when its API key is configured.

View projectInspect my contribution

My work on this project

The idea was mine. I helped build the tool, find bugs, and improve how it runs. One public contribution added Windows storage permissions and focused tests to protect saved outputs and metrics.

How it works

  1. 01

    Check the output

    Small, unsupported, or sensitive-looking results pass through unchanged.

  2. 02

    Select useful lines

    Eligible text is split into chunks. Diagnostics are preserved; Jev ranks a bounded selection when its API key is configured.

  3. 03

    Return a shorter result

    The assistant receives selected lines and an artifact ID. The supported original is stored locally.

  4. 04

    Inspect the original

    The artifact ID gives access to more detail until the stored output expires or is evicted.

Workflow checked against the project README. The linked commit verifies my Windows storage changes.

01

Engineering decision

Use Windows folder permissions and stop storage when they cannot be applied, while keeping the Unix checks.

02

What works today

The tool can reduce large outputs while keeping the original available. My linked Windows storage change passed 30 focused tests; seven existing platform failures remained outside that change.

TypeScriptNode.jsVitestWindows

Built with AlphaTales. The linked commit documents one specific contribution, not the full scope of my work. Output reduction is not guaranteed billing savings.

View projectInspect my contribution
04
Developer tool I built

Refacta

Agent-driven refactoring across a folder of files.

My part: Built the tool, SDK integration, and multi-file editingOutcome: Refactor a folder through the CLI, console, or SDK.Built by meCLI · console · SDKPublic source
Project detailsRead case studyClose case study

Developer tool

Builder · Developer tooling

Refacta is an agent-driven tool for refactoring a selected file or folder through a CLI, console interface, or SDK.

View on GitHubInspect my contribution

My work on this project

I built Refacta, including SDK integration, multi-file editing, shared-model extraction, and the scope-selection interface. The linked commit shows specific fixes within that work.

How it works

  1. 01

    Choose the scope

    Select the target files or folder through the CLI or console interface.

  2. 02

    Route the request

    Use short agent descriptions to select relevant agents and skills.

  3. 03

    Load the right tools

    Load Refacta’s own agents, with Read, Edit, and Glob tools for the selected scope.

  4. 04

    Process the folder

    Explicitly instruct the agent to find and process every matching file, rather than stop after the first.

The public commit shows my SDK, model, and console changes.

Implementation example: a shared edit model

ACTUAL CODE CHANGE · PUBLIC COMMIT

The import change

The SDK imports its edit model from the shared models module instead of the console UI.

BEFORE
from ..console.diff_viewer import EditOperation
AFTER
from ..models import EditOperation
Read the source diff
01

Engineering decision

Route the request using short agent descriptions, then load the selected agents and skills. Keep the SDK’s edit model independent of the console interface.

02

What works today

The linked commit implements two-step agent routing and explicitly instructs folder scope to process every matching file. It also fixes agent loading and the shared edit model.

PythonClaude Agent SDKTextualRichTyperPowerShell

Public source project. The code excerpt links to one of my implementation commits.

View on GitHubInspect my contribution
05
Ongoing personal product

InFit

A Flutter interface for reviewing meals and nutrition.

My part: Building the product; Flutter screens and meal-review flow so farOutcome: Meal-review screens using simulated data; Flutter prototype.Flutter prototypeIn progress
Project detailsRead case studyClose case study

Ongoing product · Interface prototype

Builder · Personal product

Explore meal logging for Indian foods through a mobile dashboard and meal-review flow.

View prototype source

My work on this project

I am building InFit as my own product. The current Flutter prototype includes the dashboard, meal review, profile, and insights, with reusable calorie and macro cards.

Illustrative InFit meal dashboard and meal review panels with sample food entries and portion controls.

UX concept · In development

Prototype flow

  1. 01

    View the dashboard

    Calorie and macro cards display sample meal data.

  2. 02

    Open meal review

    A placeholder camera action navigates to the food-confirmation screen.

  3. 03

    Adjust the meal

    Meal-review screens explore food and portion controls with simulated data.

  4. 04

    Connect real inputs next

    Camera capture, AI recognition, voice, and database saving are unfinished.

Public prototype source. This is an interface flow, not working food recognition.

01

Engineering decision

Build the interface and navigation with sample food data before connecting camera capture and AI recognition.

02

What works today

The source includes dashboard and meal-review screens. Food detection is simulated; camera capture, voice, and database saving remain unfinished.

FlutterDartGoRouter

Not a released nutrition service. Prototype data does not validate nutritional accuracy or AI recognition.

View prototype source

03 — Experience

My current role.

Jul 2024 — Present

AlphaTales

Founding Engineer / Full Stack Engineer

Brisbane, Australia (Remote)Explore the product

I joined AlphaTales in July 2024 and co-built the product with the team. I work across the flow from an idea being submitted to a plan being reviewed and exported.

  • APIs and dataBuilt FastAPI services with request validation and shared data models. My earlier database work used MongoDB and Beanie; I’m now gaining experience with PostgreSQL and SQLAlchemy following the product’s migration.
  • AI and live progressImplemented workflows with LangGraph and Azure OpenAI, including structured outputs and tool calls. The current stack uses Azure Service Bus for background messaging and Socket.IO for live progress.
  • Frontend and integrationsBuilt React screens for reviewing and exporting plans. Worked on WorkOS login, OAuth token refresh, retries, and integrations with coding assistants.
Read the project story

04 — Stack

What I work with.

Explore all 42 technologies

My main work is Python APIs, React interfaces, and AI workflows. Main tools are marked below; the full stack also includes internship and prototype experience.

Python 3.12Languages

Used for FastAPI services, Refacta, and the Azure Health agent’s monitoring scripts.

Explore AlphaTales AgentsExplore Refacta

05 — About

A bit about me.

I’m based in Chennai and work remotely with the AlphaTales team in Brisbane. As a founding engineer, I work across the product—from Python services and AI workflows to the screens people use.

I graduated in Computer Science Engineering in 2024 with a CGPA of 8.95/10. My earlier internships at LS Technologies and Nitroware Technologies covered backend services, AI chatbots, and workflow automation.

Resume GitHubLinkedIn
Premkumar A · Full Stack Engineer

TAP TO FLIP ↺

Premkumar A, Full Stack Engineer and Founding Engineer at AlphaTales. Based in Chennai. BE in Computer Science Engineering, class of 2024. CGPA 8.95 out of 10.

THE SHORT VERSION
Based in
Chennai, Tamil Nadu, India
Currently
Founding Engineer, AlphaTales
Education
BE, Computer Science Engineering

EARLIER EXPLORATIONS · COLLEGE PROJECTS

Focus

My college project used OCR to turn text in images into speech, exploring a practical accessibility use case.

Sign-language recognition

I also worked on a sign-language recognition project during college.

Certifications 5 ENTRIES

Background

Courses & certifications.

01

Generative AI for Everyone

DeepLearning.AI · Aug 2024

02

Cloud Practitioner training (AWS)

ICT Academy · Certificate of Proficiency · Feb 2023

03

Introduction to Internet of Things

NPTEL · IIT Kharagpur · Jan–Apr 2023

04

Programming using Java

Infosys Springboard · Jul 2022

05

Data Structures and Algorithms using Java

Infosys Springboard · Jul 2022

Earlier experience & education

MOST RECENT FIRST · INTERNSHIPS & DEGREE

Aug 2023 — Jan 2024

Experience

Software Developer Intern

Nitroware Technologies · Coimbatore

Built AI chatbot workflows using LangGraph and LangChain with Human-in-the-Loop validation for safer response review. Automated project setup and onboarding tasks, reducing onboarding time by approximately 25% through repeatable workflow automation.

01

Jan — Mar 2023

Experience

Python Developer Intern

LS Technologies · Coimbatore

Developed FastAPI services backed by MongoDB with structured API design, validation, and reusable backend patterns.

02

2020 — 2024

Education

BE, Computer Science Engineering

Hindusthan College of Engineering and Technology, Coimbatore

CGPA: 8.95 / 10

03
NEXT CHAPTER

Your team?

06 — CONTACT

+91 8825624586GitHubLinkedIn