[hi play above track as you read this]
i'm aryamann.
i don't just write code. i build systems.
for me, engineering isn't about memorizing syntax or debating which framework is faster. it's about seeing a complex, messy problem in the real world and designing a robust, elegant architecture to solve it.
i love the entire lifecycle of a product. the early whiteboard sketches where everything is ambiguous. the late nights debugging a race condition. the final polish before a launch. i want to understand how the database indexes, how the state flows, and how the user actually feels when they click the button.
i've found that the best way to learn isn't by reading documentation. it's by shipping.
so, let me walk you through what i've shipped.
Featured Build
Project
GUIDO
Tech Stack
Next.js, TypeScript, Tailwind, Python, FastAPI, PostgreSQL, Supabase, LLMs.
Status
Live & Scaling
The Problem
navigating massive, decentralized documentation ecosystems is incredibly painful for developers. existing search tools rely on simple keyword matching, which fails when context is required across multiple pages or repositories. i needed a way to instantly synthesize answers from dispersed engineering knowledge bases.
The Architecture
i engineered a multi-tiered rag (retrieval-augmented generation) pipeline. instead of just dumping vectors into a db, the system pre-processes docs using a specialized embedding model, clusters related concepts, and uses an llm routing layer to determine the intent of the query before retrieval even begins.
Lessons Learned
latency is the silent killer of ai products. my initial architecture took 8 seconds to return a synthesized answer. by aggressive caching, streaming the llm response, and moving the semantic search to an edge runtime, i got the p90 latency down to under 1.2 seconds. users care about speed just as much as accuracy.
building guido wasn't a straight line.
i threw away the entire codebase twice.
the first time, because the schema was too rigid to handle unstructured data. the second time, because i realized i was building for myself instead of my users.
that's the reality of startup engineering. you have to fall in love with the problem, not the code you wrote yesterday. every feature you ship is a hypothesis, and the users are the only ones who can validate it.
guido taught me how to scale, but my other projects taught me how to experiment.
opening any of those folders will show you a different chapter of how my brain works.
some projects are highly polished enterprise applications where the biggest challenge was security and state management. others are hardware integrations where i had to learn how physical sensors talk to cloud databases in real-time.
reading a textbook tells you how a system *should* work. building it tells you how it *actually* breaks in production.
but honestly? the finished projects only tell half the story.
the real learning happens in the messy, unfinished ideas.
unfinished work matters.
i use these "open builds" as a sandbox to test emerging technologies without the pressure of scaling them into businesses.
it's where i break things on purpose. it's where i test if the latest research paper actually holds up when you try to write the code for it.
learning in public through these experiments has shaped my foundational engineering toolkit.
Engineering Toolkit
Programming Languages
JavaScript, TypeScript, Python, C++, SQL
Frontend Ecosystem
React, Next.js, Tailwind CSS, Framer Motion, Redux, Zustand
Backend & APIs
Node.js, Express, FastAPI, Django, GraphQL, REST
Databases & Storage
PostgreSQL, MongoDB, Redis, Supabase, Pinecone (Vector)
AI & Machine Learning
LangChain, LlamaIndex, OpenAI API, HuggingFace, PyTorch
Cloud & Infrastructure
AWS, Vercel, Docker, GitHub Actions, Nginx
Design & Hardware
Figma, Arduino, Raspberry Pi, IoT Sensors
technologies come and go. frameworks deprecate. paradigms shift.
but the core principles of engineering—understanding trade-offs, designing resilient architectures, and building things people actually want to use—remain exactly the same.
that's why i write down everything i learn.
RESOURCE VAULT
[ Development Resources ]NOTES FROM THE ENGINEERING
Slide to explore right →if you want to build something incredible together, let's talk.
