Anushka Joshi — machine learning and software engineer.
I'm Anushka. I build software that solves real problems — machine learning models, intelligent automation, and full-stack applications.
I like the part where a research idea turns into something people can actually open and use.
I'm mostly interested in how things actually work.
Curiosity comes first. I want to know why something behaves the way it does before I have an opinion about it. Most problems get smaller once you look at them closely.
Small pieces, plain code, one clear idea at a time. I'd rather write something a teammate can read on a Monday than something clever I'd have to explain.
It's the most interesting place to build right now — the research moves fast, but the useful part is turning it into software people can rely on. I like sitting between those two.
I ask a lot of questions early, then get quiet and build.
I try to keep experiments cheap and the things people depend on boring. Reliability is a feature; so is being able to change your mind quickly.
- Multi-agent systems
- Long-context RAG
- AI evaluation
- AI infrastructure
- Human-AI interaction
Things I've made, and what they were for.
Every one of these is a real repository. Open a category to look inside.
CodePilot
Three small agents — a planner, an architect and a coder — that hand work to each other in stages and return a downloadable project.
Salesforce & API Automation Workflows
Event-driven workflows and API integrations that keep Salesforce in step with the marketing and analytics tools. Built with Python, webhooks, Zapier and REST APIs. (Built during my Mailmodo internship.)
Intelligent Invoice Extraction Pipeline
OCR feeding a language model, with validation and a fallback path, writing tidy records into SQLite behind a small dashboard.
Where I've worked, and what came of it.
Customer Solutions Engineer — Intern
@ MailmodoI wrote the code that sat between Salesforce, the marketing tools and everything internal. Small scripts, mostly. They ran every day on real customer data, which is a good way to learn what breaks.
- Kept Salesforce leads and contacts in sync over REST APIs and webhooks, so nobody had to retype anything.
- Tidied up workflows that spanned several platforms and made the data between them agree.
- Wrote Python, JavaScript and Apps Script jobs to replace recurring manual work.
- Sat in on client calls, then went and built what came out of them.
Where I learned the basics, and kept going after.
Things I reach for often.
I write things down until they make sense.
Most of these started as notes to myself — something I'd half-understood and wanted to explain properly. They live on Medium.
Business Automation Evolved: GenAI vs Agentic AI
GenAI dreams. Agentic AI delivers.
Concept to Code: Master Top ML Algorithms
Decode. Build. Deploy.
Machine Learning, But It's Not That Deep (Yet)
ML Algorithms 101 — what they are, how they work, and where to use them.
Your First Brain Dump on Machine Learning
From Huh? to Aha! — your intro to ML.
Working on something interesting?
I'd genuinely like to hear about it — a product, a stubborn bug, a half-formed idea. I read everything that lands here and reply to most of it.