Abhijith N M
@abhimanjakal_1d056c29
No bio provided.
About Abhijith
Founder, writer and editorI'm a software developer based in India. I work on web applications, AI features and the infrastructure around them.
I started nexGenCoders because the resources I wanted while learning were scattered: a tutorial that skipped the setup, a course that was out of date, an AI answer that solved the exercise without explaining it. The site is my attempt to put the explanation, the diagram and a runnable example in one place.
I write and review most of the lessons here, run the code samples, and answer the correction emails. If something is wrong, tell me and I will fix it.
Articles by Abhijith
A small team's AI eval stack
You do not need an eval platform. A folder of cases, a pytest runner, a rubric and a CI gate will catch most regressions.
Abhijith N M
Prompt injection is an architecture problem
A model cannot reliably tell instructions from data. If an agent can read untrusted text and take actions, the fix is permissions and boundaries, not a stronger warning in the prompt.
Abhijith N M
Context engineering is mostly housekeeping
The model only sees what you put in the window. Ordering, compression, delimiting and caching matter more than another clever instruction.
Abhijith N M
Why your AI feature feels slow, and how to fix the latency budget
Users judge the tail, not the average. Break the AI request into a latency budget, then attack the largest span instead of guessing at the model.
Abhijith N M
What to learn as a junior developer when AI writes code
AI can produce a working first draft. It cannot tell you why the draft is wrong, which trade-off will hurt in six months, or how to debug a production incident. Those are the skills to build.
Abhijith N M
Six numbers to check before you ship an AI feature
Cost per successful task, p95 latency, token budget, cache hit rate, eval pass rate and tool failure rate. If you cannot state these, you are not ready to ship.
Abhijith N M
RAG after the hype: the retrieval patterns that still hold up
Naive vector search demos well and fails quietly. Structure-aware chunking, hybrid retrieval, permission filters and retrieval evals are the parts that survive contact with real documents.
Abhijith N M
MCP vs function calling: when you actually need a protocol
Function calling connects a model to your code. MCP connects a model to tools it did not ship with. The difference matters when you have more than one host or more than one team.
Abhijith N M
AI agents in production: the boring parts that decide whether they work
A working demo is the easy part. State, retries, permissions, evals and cost controls are what separate an agent that survives production from one that only works in a notebook.
Abhijith N M
5 Udemy Python Courses Worth Your Money in 2026 (A Student's Honest Take)
Confused about which Python course to pick on Udemy? I went through the most popular ones so you don't have to. Here are the 5 best Python courses in 2026 that are perfect for students — even if you've never written a single line of code before.
Abhijith N M