About
I’m Pratham, a software engineer in Pune, India. I finished a B.Tech in Computer Science at MIT-ADT University in 2026.
My work spans three layers that usually belong to different people: the product a user touches, the cloud infrastructure under it, and the AI or ML that makes it smarter. I started in research, classifying EEG signals for a prosthetic arm. Then came products and client work, and in 2026 production cloud engineering on AWS and Google Cloud.
What ties it together is an interest in how systems behave: where the data goes, what it costs, what breaks first. The journey page has the longer version.
How I use AI in engineering
AI makes me faster at the parts of engineering that are search, recall and first drafts. It doesn’t make decisions for me.
AI multiplies judgment; it doesn’t replace it
I use AI assistants for code exploration, debugging, test scaffolding, documentation and first drafts of repetitive code. Architecture, security boundaries and what ships stay my decision, and generated code gets the same review as any other contributor’s.
Write the plan before the code
Larger changes start as a short spec and an implementation plan checked into the repo. The Force Dark Mode repository keeps these under documentation/specs and documentation/plans. An AI pair works far better against a written plan than against a vague prompt.
Verify claims against the system
Generated answers get checked against the code, the docs or a running system before I rely on them. The same rule applies to this site: every number on it names its source.
Measure cost and performance, then change things
The Searce FinOps work and the Abhinandan Mountreea performance pass both started from measurement: billing and utilization data in one case, load time in the other.
Where it earns its place
- Code exploration
- Mapping an unfamiliar codebase before changing it, e.g. tracing a request path across Next.js, Lambda and DynamoDB.
- Debugging
- Forming and ranking hypotheses from stack traces and logs, then confirming each one by hand.
- Tests and fixtures
- Scaffolding test cases and edge cases, then trimming them to the ones that actually guard behaviour.
- Documentation
- Drafting READMEs, architecture notes and compliance docs from the code, then editing for accuracy.
- Prototyping
- Getting a throwaway version running quickly to test an idea before designing the real one.
- Review
- A second pass over diffs for missed error handling, injection risks and unclear naming.
Setup
Hardware
- MacBook Air M3
Editor and terminal
- VS Code
- Cursor
- GitHub Copilot
- zsh + Git
Build and ship
- Node.js + pnpmPackage manager across these repos
- DockerStreamVault progress API
- AWS CDKStreamVault infrastructure
- VercelFrontend hosting, including this site
Cloud and data
- AWSS3, CloudFront, Lambda, API Gateway, DynamoDB, Cognito, EC2, SES
- Google CloudAssociate Cloud Engineer
- SupabasePostgres + pgvector
- MongoDB
AI
- ClaudePair programming, review, docs
- OpenAI, Gemini, Groq APIsInsightVault embeddings and generation
Common questions
Who is Pratham Shah?
Pratham Shah is a 2026 Computer Science graduate (B.Tech, MIT-ADT University, Pune) and a Google Cloud Associate Cloud Engineer. He works on cloud, backend and full-stack engineering.
What cloud experience does Pratham have?
As an Associate Cloud Reliability Engineer intern at Searce (Apr–Aug 2026) he analyzed Google Cloud billing and utilization data and prepared rightsizing and cost-optimization recommendations. His StreamVault project defines Cognito, CloudFront, Lambda and API Gateway infrastructure in AWS CDK.
What did Pratham do at NeuraMach AI Studios?
He was an AI/ML Full-Stack Engineering Intern with a QA focus (Jan–Apr 2026, Pune). He built Next.js and TypeScript interfaces using SSR and SSG, integrated FastAPI REST endpoints with React Query and Axios, and used AI coding tools such as GitHub Copilot and Cursor.
What AI work has Pratham done?
He built InsightVault, a question-answering app over uploaded PDFs using Supabase Postgres with pgvector, multi-provider embeddings and grounded LLM prompts. Earlier he worked on EEG classification for the NeuroArm brain-computer interface team project.
What roles is Pratham looking for?
Junior cloud, backend, FinOps and full-stack engineering roles.
How can I contact Pratham?
Email pshah88669@gmail.com, or connect on LinkedIn (linkedin.com/in/pratham-shah-729432258) or GitHub (github.com/pshah-lab). A resume is at pshah.fun/resume.