Greater Noida, India — backend & AI engineer

Md AnishAhamad

I build backend services and agent systems, and I spend most of my time working out where AI-generated code goes wrong. That second part is my actual job — at Handshake I write the tasks that test coding agents, then grade what they produce.

What I've shipped Resume GitHub LinkedIn
incident-engineer · 11 real incidents · same verifier for both runs
two agents, with a verification gate
11 / 11 fixed
a single prompt, no gate
0 / 11 fixed
The difference is one rule: a fix is only accepted once the full test suite passes. Without it the agent writes patches that read perfectly and do not work. Read the code →
01 / shipped

Two products, both live and in use

Not side projects. Real users, running right now, maintained by me.

exactsizetool.com

live

Indian government exam forms demand an exact photo size, an exact signature ratio and a hard KB limit. Get it wrong and the form is rejected. This fixes that — around ninety free tools for images, PDFs and documents.

  • Everything runs in the browser. Your photo and signature never leave your device, so there is no server for me to leak.
  • An assistant reads the raw instructions off an exam portal and sets up every tool for you.
  • No account, no daily limit, no watermark. Paid for by ads.
90+tools
0files leave your device
Open the tool →

gateguide.co.in

live

A GATE preparation business I built and run. Six books — Computer Science, Electronics, Civil, Data Science & AI, Electrical and Mechanical — written by counting real question papers one at a time, so the weighting follows what is actually asked rather than the syllabus order.

  • Every question comes with a worked solution, not just an answer key.
  • Ten full mock tests in each book, built from the same paper analysis.
  • Written, published, priced and supported entirely by me.
6books
5,732questions
60mock tests
80free guides
Visit →
02 / work

Paid to find where models are wrong

Every role I have held asks the same question: the model produced something — is it actually correct, or does it only look correct?

Jul 2026 — now

AI Coding Task Author

Handshake

I write realistic software-engineering tasks used to test AI coding agents — the issue description, a reference solution, and the automated tests that decide whether a fix is correct. Then I grade what the agents produce. Reading AI-written code and deciding whether it is right is the whole job.

Jul — Sep 2026

CFD Engineer / AI Task Designer

Alignerr

Designed AI evaluation tasks in computational fluid dynamics — precise problem statements, reference solutions and deterministic checking, using OpenFOAM.

Jan — Apr 2026

LLM Post-Training Intern

Ethara AI

Post-training work on large language models: refining datasets and engineering prompts to improve output, and analysing the cases where the model got it wrong to judge whether it was ready to ship.

2022 — Jun 2026

B.Tech, Computer Science & Engineering

NIET Greater Noida · 8.42 CGPA, Honors

Built and shipped both products above alongside the degree.

03 / code

Things I've built

The first three are the ones I would stand behind in an interview. The rest are smaller, and I would call them prototypes.

incident-engineer python

Two agents that find and fix real bugs. A triage agent forms a theory from the logs; an investigator uses tools — read, grep, run-tests, sandboxed Python — to find the cause and patch it. A fix only counts once the tests pass. 11/11 against a baseline's 0/11.

↗

KaamSathi django · react

Full-stack team task manager. Django REST and PostgreSQL behind JWT auth, a React SPA in front, and role-based access at two levels — platform-wide and per project. Deployed on Railway.

↗

ai-resume-generator java · react

Spring Boot API that turns a prompt into a structured resume — clean JSON, not prose, so the front end can render it reliably. The model backend is pluggable, so OpenAI, DeepSeek or a local model all work behind one interface.

↗

mermail-expense-tracker mcp skill

Turns receipt emails into an expense ledger over MCP. Treats every email as untrusted input, has no access to payment tools, and never sends anything without a human approving it.

↗

cloud-native-devops terraform

Terraform for an AWS VPC, subnet, security groups and EC2; a Dockerised app behind Nginx; GitHub Actions building and deploying on every push.

↗

api-etl-pipeline fastapi

Pulls records from external REST APIs, normalises inconsistent JSON into one schema and loads it into PostgreSQL. Reusable connector logic, retries and failure tracking.

↗

everything else github

A RAG service, a LangGraph multi-agent workflow, a blockchain dashboard built for a bounty, and a few smaller experiments.

↗
One I am still building: an agent that finds roles, scores them against what I have actually done, and tailors a resume per role. I set out to make it apply automatically — then I checked, and six of seven real application forms sit behind a CAPTCHA. So it now does everything up to the submit button, and I press it. Automate the work, keep the verification.
04 / stack

What I work with

languages
PythonJavaSQLJavaScriptBash
backend
FastAPISpring BootDjango RESTREST API design Spring Data JPAJWT auth
data
PostgreSQLMySQLMongoDBETL
infrastructure
DockerGitHub ActionsTerraformAWS EC2 AWS VPCNginxLinuxGit
ai & agents
AI evaluationBenchmark authoringAgent architecture Tool-calling loopsLangGraphMCPRAG FAISSStructured output
05 / contact

Let's talk

Open to backend and AI engineering roles. Based in Greater Noida, happy to relocate.

[email protected] email linkedin.com/in/md-anish-ahamad linkedin github.com/anishraja04 github resume open