Founding AI Engineer @Superagent, building production AI systems that actually work.

I started with data science, practiced ML and deep learning, then fell in love with AI agents and backend systems. There's something deeply satisfying about building things that actually work in production: seeing your code handle real traffic, solve real problems, and survive real failures as opposed to just seeing the loss function go down. (Sorry ML nerds)

I went from intern to Founding AI Engineer in 6 months. I own systems end-to-end: translating "we need this vague-feature-description" into actual technical specs, designing the system, building it, deploying it, and then maintaining it. The full lifecycle experience™.

I also have 10+ certificates from DeepLearning.AI, Imperial College, Khan Academy for Machine Learning, Deep Learning and Math. They're worth approximately as much as the paper they're not printed on. 📜

multi-agent systems: building and keeping up with the ever-evolving world of LLM agents, architectures, and agent skills. LangChain, LangGraph, LlamaIndex, Google ADK, GenAI SDK, vector search with ChromaDB. I've touched enough agent frameworks to have opinions about all of them.

backend done properly: FastAPI services with proper error handling, Celery workers and Redis for task queues, rate limiting that actually limits, PostgreSQL/Supabase schemas that make sense. Integrating with APIs that have questionable documentation (WhatsApp, LINE, Messenger, you know the ones). Making sure webhooks actually webhook.

production ops: deployments on AWS/GCP/Azure, NGINX configs, SSL certs that don't expire at the worst possible moment, database migrations that don't nuke your data, monitoring that pages you before users complain. The unsexy stuff that keeps everything running.

debugging grind: tracing issues through logs, databases, and third-party APIs until 2am because something broke and it's definitely not my code (it was my code).

These are from over a year ago. Could probably redo any of them in a day or two now, but they were good learning experiences:

AI Interview Agent: Real-time voice-based interview system with WebSockets, STT/TTS integration, and multiple collaborating agents (Interviewer, Coach, Skill Assessor). FastAPI backend, RAG and web-search. The frontend? Let's just say it works and we don't talk about how (vibecode go brrrr). 🚀

Transformer from Scratch: Implemented the full "Attention Is All You Need" architecture in PyTorch for English-Vietnamese neural machine translation. Multi-head attention, positional encoding, the whole paper. I took a seminar about this paper in class, so I was like "might as well implement it while I'm at it" 🤷

MLOps Pipeline: Crypto volatility prediction with XGBoost, deployed on AWS EC2 with cron-automated training, FastAPI serving, and proper logging. Because models that only work in notebooks don't count 😉

Sudoku Game: Full-featured game with user auth, leaderboards, hint system, and game saving. Flask backend, Tkinter frontend, SQLite. Yes, I built a GUI app. No, I don't want to talk about Tkinter 🙅

...and a few more things like a document denoising autoencoder, blog platform with AI moderation, etc.

"Founding AI Engineer" sounds fancy, but honestly it reflects breadth more than depth. I've touched a lot of things, been part of a lot of decisions, and gotten my hands dirty across the stack. Grateful for the exposure and the chance to build and own production systems end-to-end this early.

Still very much learning. It's a journey.

Also actively working on the fundamentals: SOLID, DRY, YAGNI... I follow them but don't chain myself to them. Trying to write more modular, maintainable code. Looking to go deeper on system design and building things that are actually a joy to maintain six months later.

If any of this resonates, hit me up:

• 🏆 Hackathons. Always down for a good one.
• 🌐 Open source. Looking to start contributing more.
• 🎮 Researching how LLM agents can learn transferable strategic skills across game-theoretic environments using judge-guided feedback (WIP)
• 🦞 OpenClaw. Currently obsessed with tweaking its personality and watching it do ridiculous things autonomously. The space lobster is real.
• 💬 Or just about anything. Don't be shy.

• 📧 Email: ranjitn.dev@gmail.com
• 💼 LinkedIn: linkedin.com/in/ranjit-n