Balaji Arumugam

Software Engineer building AI products and distributed systems.

  • • Hey there! I'm Balaji, a 25-year-old who's been glued to screens since I could walk.
  • • Currently grinding through my Master's in CS at Georgia Tech (ML track, because who doesn't want to teach machines to think?).
  • • By day, I'm building conversation intelligence platforms at Observe.AI, making customer service smarter using AI.
  • • By night (and weekends), I'm an indie hacker shipping products that solve real problems - from AI-powered creativity tools for kids to resume roasting bots that don't hold back.

Slide into my DMs: b0a04gl@proton.me | Call me maybe: +91 8056441286 | My corner of the web: hn/b0a04gl


What I'm Up To

SDE2 @ Observe.AI (aka the place that pays my bills)

May 2024 – Still here, still caffeinated

Observe.AI Platform

The AI platform I help build at Observe.AI

I work on the conversation intelligence platform that powers autonomous contact centers. We're using AI to transform how customer service works - automating calls with VoiceAI, providing real-time assistance to agents, and handling quality assurance automatically.

  • Built data export platform enabling enterprise customers to extract conversation analytics to their data warehouses (Snowflake, S3) with custom scheduling and transformations -- achieved 90% QoQ customer adoption growth while reducing warehouse costs by $60K/year through query optimization.
  • Re-architected API gateway with Kubernetes service discovery in Kong, eliminating external DNS/ALB hops -- reduced P95 latency by 55% (564ms→248ms) and hardened security by keeping internal traffic within EKS cluster.
  • Built intelligent filter extraction and conflict resolution system for NL-to-SQL queries using disjoint-set union algorithm -- solved ambiguous entity parsing ('@filter1' syntax), aggregation vs. projection detection, and sequence ordering with client-side in-browser DB persistence for secure, performant multi-source data filtering.
  • Engineered observability migration from NewRelic to VictoriaMetrics with agentic monitoring system (80% AI, 20% human-in-the-loop) for contextual alerting -- contributed to team reduction of incidents by 87% (15→2/quarter) with avg MTTD <15min, avg MTTR <100min for past quarter

Indie Hacker (aka nights and weekends)

📊 View Live Dashboard

Forever – Because sleep is overrated

When I'm not at my day job, I'm building products that scratch my own itch. I love the entire process - from identifying problems to shipping solutions that people actually use. There's something magical about going from zero to one without corporate bureaucracy slowing you down.

My approach is pretty straightforward: build fast, get real feedback, iterate quickly. No fancy business plans or pitch decks - just solving problems for real users and seeing what sticks. So far, it's led to ccplay.com, ahaharesume.com, and fpaisa.sh - each one teaching me something new about product development, marketing, and what people actually want.

The indie hacker community has been incredibly supportive. There's something powerful about a group of people all building in public, sharing wins and losses, and helping each other grow. It's the perfect balance to my day job - complete creative freedom with direct user feedback.

Current stats: $200-$450 MRR, 15,650+ total MAU across projects. Check the live dashboard →

Grad School Grind

Georgia Tech | Jan 2025 – Present

MS in Computer Science, ML track (because neural nets are cool)


Cool Stuff I Built

cocomelonplay.comAI creativity platform for kids (ages 5-10)

Built co-drawing, interactive lessons, and story generation features. Measured 30% fluency, 56% narrative cohesion, and 12% originality improvements in user-generated content.

Stack: Next.js, TypeScript, GCP, LaTeX, Gemini Imagen, Shadcn/ui

Status: 650+ MAU, $200-$450 MRR, MVP built in 3 weeks

ahaharesume.comResume formatter + AI roasting tool

Resume formatting and AI-powered feedback tool. Launched in 2 days and scaled through organic word-of-mouth.

Stack: Next.js, TypeScript, GCP, LaTeX, GPT-4o, Shadcn/ui

Status: 15K+ MAU, $200+ in donations, built in 2 days

ballgorithmUnbeatable ping-pong AI in 80 lines

Fully code-golfed JavaScript ping-pong game with unbeatable AI. Built for fun and learning.

Stack: JavaScript, Canvas API

Status: 40K+ impressions on social media, built in 2 hours

kids playdateParent-driven playmate matching platform

Matches kids into groups of 4-5 based on weighted parent preferences and child interests for shared activities.

Stack: Next.js, TypeScript, Shadcn/ui

Status: Ongoing development


Random Thoughts

Typing 118 WPM Broke My Brain in the Right Ways (A Year-Long Journey)

March 2025 • 8 min read • Backend

How I went from 60 WPM to 118 WPM and accidentally rewired my brain for focus and how i got viral on Hacker News. Spoiler: it wasn't about the home row.

Built a Kids' App, Hit $900 MRR, Almost Burned Out (A 12-Week Reality Check)

February 2025 • 10 min read • Startup

The unfiltered story of building CCPlay from 'random idea' to actual revenue. Spoiler: kids are way cooler than adults.

Stop Letting ChatGPT Write Your Thoughts (An Authenticity Rant)

January 2025 • 6 min read • Opinion

Hot take: AI-generated content is making everyone sound the same. Here's why your messy, imperfect thoughts are infinitely more valuable.

View all posts →


My Journey

SDE2 (got promoted from SDE1)

Walmart Global Tech | Jan 2021 – May 2024

  • Saved the company $40k/year by optimizing cloud resources (turns out running everything at max capacity isn't the way)
  • Led a team of 3 devs to integrate LG and Samsung for some "Future Inventory" thing that made $2M. Not bad for a Tuesday
  • Boosted dev productivity by 10% across 12 teams with a multi-tenant audit framework (because who doesn't love good logging?)
  • Made a critical API 25% faster by switching to gRPC (REST is so 2020, amirite?)

Intern (aka coffee fetcher)

Samsung R&D | May 2020 – Aug 2020

  • Profiled SmartThings mobile app performance, identified render bottlenecks -- reduced initial load time by 200ms through code-splitting and lazy-loading of non-critical UI components.

My Arsenal

Languages I Actually Use

Python (my first love), Java (for when I hate myself), JavaScript, TypeScript (JS but better)

Frameworks That Don't Suck

Spring (enterprise stuff), Next.js (React but easier), FastAPI (Python web done right)

AI/ML Toys

PyTorch, Hugging Face (transformers everywhere), vLLM, MCP, NumPy, Pandas (data wrangling)

Cloud & Data Stuff

AWS (when I have to), GCP (my preference), Snowflake, MongoDB, Elasticsearch

Tools I Live In

Git (obviously), Docker (containers ftw), Vim (fight me), Cursor (AI coding), asyncio, Pydantic, shadcn/ui


School Stuff

Master's in Computer Science

Georgia Institute of Technology | Jan 2025 – Present

ML track because neural networks are the future (or so they tell me)

Bachelor's in Information Technology

Anna University, Chennai | June 2017 – Jan 2021

AI specialization (was into AI before it was cool)


Hit Me Up

Drop Me a Line