STSridharan T
CASE NO. ST-2026Open to product roles
Evidence verified

Software engineer · Chennai / Salem, India

I build thesystems behindintelligent products.

I'm Sridharan — Software Engineer at Aspire Systems. I help build a campaign platform serving 50K+ users and ship open-source AI infrastructure on the side.

The thread: every project is a constraint made legible — then shipped as something another person can run, inspect, and maintain.

Examine the evidenceMethod
GitHub LinkedIn

Evidence locker

Pull an exhibit.

Social Media Agent — LangGraph

EXHIBIT A

Agent routing

One campaign idea becomes four platform-specific drafts through a stateful workflow with memory and retries.

Input

One idea

System

LangGraph

State

Memory

Output

4 drafts

4 parallel branches · retry on failure

Finding: Open source · parallel branches · checkpointed

LangGraphOpenAIAnthropic Source

Showing Agent routing: Open source · parallel branches · checkpointed.

Evidence, not theatre — click only.

Filed outcomes — cross-referenced to experience log

01 / 03

50K+

active users on a campaign platform I help build

View in log 02 / 03

40%

lower API latency via MongoDB + pipeline tuning

View in log 03 / 03

30%

lower media-storage cost through S3 workflows

View in log
Evidence Index03 / 09 files

Proof over promises.

Three builds, three constraints: agent orchestration, cognitive-load-aware planning, and reviewable AI edits inside the editor. Each file keeps its outcome on the front.

FILE-01012025 · Open source

Social Media Agent

“One graph. Four platform-specific draft flows.”

A LangGraph workflow that turns one campaign idea into platform-specific drafts for LinkedIn, Twitter/X, Facebook, and Instagram—adapting tone, length, and hashtag strategy for each channel.

  • Orchestrated parallel platform branches via LangGraph StateGraph with memory checkpointing
  • AI rewrites content per platform: 280-char Twitter punchy vs long-form LinkedIn professional
PythonLangGraphLangChainOpenAIAnthropicDocker
Source
FILE-02022025 · Prototype

FocusFlow — AI Task Manager

“Context-aware scheduling designed around cognitive load.”

A task-planning PWA exploring cognitive-load-aware scheduling. Google Genkit prioritizes tasks using current time and workload context, with real-time push notifications on iOS.

Next.js 15FirebaseGoogle GenkitPusher
Source
FILE-03032025 · Open source

AI Code Assistant — VS Code Extension

“Keyboard-first AI edits with reviewable inline diffs.”

A VS Code extension with Ghost Chat—a cursor-positioned AI panel that previews inline diffs before any change is applied.

TypeScriptVS Code APIWebViewMonaco Editor
Source

Archive

Inspect the full archive — 9 case files, each with stack, highlights, and source. ONNX lives as a dedicated build x-ray on this page.

Open the archive
Build X-Ray — Dossier exhibit

An AI model you can carry in one file.

FILE: ONNX-2025 — sentence-transformers/all-MiniLM-L6-v2

I turned an embedding model into a portable service that runs without Python or a network — and ships across three operating systems. This is the one place ONNX appears on the homepage.

Inspect the source

Chain of custody — request → delivery

EVIDENCE BAG — ONNX

22MB

Model + tokenizer + runtime + API

all-MiniLM-L6-v2 · 384 dims

LINUXWINDOWSmacOSNuitka

— X-ray: one binary, zero setup —

  1. 01

    Request

    POST /v1/embeddings

    Plain text enters a small, predictable REST contract.

  2. 02

    Inference

    ONNX · MiniLM-L6-v2

    The bundled model returns a 384-dimensional embedding — fully offline.

  3. 03

    Package

    Nuitka standalone

    Model, tokenizer, runtime, and API become one 22MB binary.

  4. 04

    Deliver

    3 operating systems

    GitHub Actions matrix builds Linux, Windows, macOS; Docker stays available.

✓

Offline

No network needed

✓

Zero setup

No Python install

✓

22MB

Standalone binary

✓

384

Embedding dims

All facts from repo: 22MB · 384 dims · Nuitka · GitHub Actions matrix (Linux/Windows/macOS) · Docker

Career log

From trainee to production ownership.

A steady progression at Aspire Systems — from learning the stack to owning systems that carry real users, latency budgets, and storage costs.

Ledger — Aspire Systems

3 entries

  1. Apr 2024 — Present · Chennai

    Software Engineer

    Production platforms, AI integrations, and delivery ownership.

    Current — primary file

  2. Jan 2024 – Mar 2024

    Trainee Engineer

    Contributed backend service development in Node.js and Laravel for production platform features

  3. Jan 2023 – Mar 2023

    IMPACT Trainee

    Completed structured full-stack training covering JavaScript, Node.js, MongoDB, and REST APIs

Each entry retains its own achievements — no “3+ years” compression.

Active assignment

Software Engineer

Aspire Systems · Chennai · Full-time

Apr 2024 – Present

Core product flows

Campaign management, analytics, delivery tracking, and rich-text email creation.

System performance

MongoDB query tuning, response pipelines, and media-storage paths at 50K+ scale.

Applied AI

Content moderation, generation workflows, and citation-aware realtime chat.

Delivery ownership

Reusable npm packages, WebSocket interfaces, GitHub Actions → AWS EC2.

50K+

users served

40%

latency ↓

30%

S3 cost ↓

The common thread is end-to-end responsibility: follow the behavior from the interface through the API and data layer, then carry it into production.

Detailed achievements per role preserved — see source in data.js

Capabilities — no percentages

From constraint to working software.

A stack list says what I've touched. This shows how I use it: frame the real constraint, connect the system, design the recovery path, and close the delivery loop.

Method — 4 phases · dossier tabs

Phase 01 — Frame

Name the real constraint.

“What must become measurably better?”

I start with the user path, the system boundary, and the signal that will tell us the work actually helped.

Filed output

  • Success measure
  • Failure boundary
  • Smallest useful slice
Tools in this phaseSystem mappingAPI contractsProduct thinking

Showing Frame: Name the real constraint.

Working stack — grouped by evidence

01

Production core

TypeScript · Node.js · MongoDB · Laravel

02

Applied AI

Python · LangGraph · OpenAI · ONNX

03

Product interface

React · Next.js · Vue · Tailwind CSS

04

Delivery

Docker · AWS · GitHub Actions · Vercel

Dossier — principal

Curious enough to explore. Practical enough to ship.

I'm a Salem-based software engineer at Aspire Systems. At work, I help ship an email-marketing platform used by 50K+ people; outside work, I build open-source AI infrastructure and developer tools.

I'm drawn to problems that cross boundaries — model behavior, API design, performance, packaging, and UX. My best work begins with a measurable constraint and ends with something another person can run, inspect, and maintain.

“The work counts when someone else can run it without me in the room.”

— Working style, filed
Salem, Tamil NaduChennai — Aspire SystemsBannari Amman Institute of Technology ’23