Ibu's résumé

with receipts

Mohamed Ibrahim A.

Software & AI engineer · Coimbatore, India

Printed 1 November 2026 · 14 claims, each with a receipt

  • Verified every line has public proof
  • Partial some lines are on my word
  • Self-reported no public record yet

Profile

Final-year B.Tech Information Technology student in Coimbatore, graduating in June 2027. I build AI agents and serverless apps on AWS, often in public and often against a weekend deadline.

My one rule: the model never invents a fact. This résumé keeps it too. Every claim links to its receipt, and every receipt prints its catch, the part that makes me look worse.

  • Ground the model. Facts come from data and tools, never from the model's memory.
  • Publish the number that hurts. An R² of 0.94 means little without the 47.93% beside it.
  • Own the loop. Fifty lines I can debug beat a framework I can't.

Experience

Selected work

  • #1 CryptoPulseVerified

    R² 0.94 on Bitcoin close price · Infosys Springboard internship

    • R² · close · test set0.9402
    • Direction · close · test set47.93%

    The catch: On close price the model calls up-or-down correctly 47.93% of the time.

    Corrected: An earlier write-up of mine called this an 80/20 split. The LSTM cell in the notebook splits 70/30.

    Evidence: Notebook: “Evaluating Model” cell

  • #2 Supermarket Ops AgentVerified

    30 typed tools, one chat window · hiring take-home · 2026

    • Tools in ALL_TOOLS30
    • Where prices come fromPostgres, via tools

    The catch: One store per bot.

    Corrected: My README says 32 tools, and so did an earlier write-up. The registry holds 30. The code wins.

    Evidence: registry.ts: the 30 tool schemas · 3 more on the receipt

  • #3 DocuMind AIVerified

    Built in a weekend on Amazon Bedrock · AWS Builder Showcase Challenge · Aug 28–31, 2026

    • ModelAmazon Nova Pro
    • BackendLambda · Python 3.12

    The catch: There is no retrieval.

    Evidence: Published write-up on AWS Builder Center · 2 more on the receipt

  • #4 TripwireVerified

    A ~0.2 s reflex against prompt injection · Strands Agents + Amazon Bedrock · Oct 2026

    • Attacks that worked · Tripwire0 / 10
    • Median decision · guard233 ms

    The catch: End to end, the median guard decision took 341 ms, not 0.2 s.

    Evidence: README: results tables · 2 more on the receipt

  • #5 Ghost ProbeVerified

    A read-only S3 tool that found “deleted” data · AWS Builder Center “Give your Agent a Tool” challenge · Oct 2026

    • IAM actions granted11, all List/Get
    • Verdict on case_25 · beforeCOMPLETE
    • Verdict on case_25 · afterINCOMPLETE

    The catch: That's one case from a fictional dataset, run against a demo bucket I seeded myself.

    Evidence: Least-privilege IAM policy · 2 more on the receipt

  • #6 The Git BardVerified

    A bard that reads my commits every Friday · AWS “Set Your Creative App Free” weekend challenge

    • EventBridge schedulecron(0 16 ? * FRI *)
    • StorytellerAmazon Nova Micro

    The catch: It only chronicles the repos I list (mine lists one), and in the SES sandbox it can only email addresses I've verified.

    Evidence: template.yaml: the Friday schedule · 1 more on the receipt

  • #7 Voter SaathiVerified

    An election guide in 12 Indian languages · civic tech · 2026

    • Languages12
    • Models tried, in order3 free OpenRouter models, then Gemini

    The catch: Only English is hand-written.

    Corrected: My README credits Gemini with the other 11 languages. The server tries three free OpenRouter models first (Gemma, Qwen and Llama) and calls Gemini only if they fail or no OpenRouter key is set.

    Evidence: README: the 12 languages · 2 more on the receipt

  • #8 CelebrAIVerified

    86 passing tests, zero runtime dependencies · event planner

    • Unit tests passing86 / 86
    • Runtime dependencies0

    The catch: All 86 tests passed while a dialog was painted over the entire app.

    Evidence: package.json: dependencies {} · 2 more on the receipt

  • #9 Tiny Tale ToyVerified

    Every story branch generated live · Bedrock + Lambda + Amplify

    • PathAPI Gateway → Lambda → Bedrock
    • Default model in codeMinistral 3 8B

    The catch: Models don't always return clean JSON.

    Corrected: The README still says Nova Lite in places. The code's default model is Mistral's Ministral 3 8B.

    Evidence: lambda_function.py: the defensive parser · 1 more on the receipt

Skills

  • Agents that don't make things up Amazon Bedrock · Strands Agents · Claude API · Gemini API · LangChain · RAG
  • Serverless on AWS Lambda · API Gateway · EventBridge · S3 + CloudFront · SES · SAM · Amplify · IAM
  • Money-grade full-stack TypeScript · React · Node + Express · FastAPI · PostgreSQL + Prisma · MongoDB · Oracle APEX
  • Honest ML TensorFlow / Keras · scikit-learn · Computer vision · NLP
  • Also fluent in Python · JavaScript · TypeScript · SQL & PL/SQL · Docker · GitHub Actions · Google Cloud Run · Firebase
  • Studied, not yet shipped Java · C · C++ · Kubernetes · Azure · BigQuery · Snowflake

Community

Education

  • B.Tech, Information Technology · CoimbatoreJune 2027

This portfolio

Totals

  • Claims on this résumé14
  • Backed by public proof10
  • Partly proven3
  • Self-reported1
  • Corrected by my own code4
  • Invented facts0