Muhammed Faiha

Muhammed Faiha

Backend and AI Engineer

I build backend systems that stay correct under load, and AI pipelines that stay fast under real traffic. Working on LLM services and retrieval infrastructure.

View Resume GitHub LinkedIn
02

About

I am a final year Computer Science Engineering student at SRM Institute of Science and Technology, my work mainly revolves around VLMs, LLMs and SLMs. I also enjoy working with RAG, agents and retrieval infrastructure.

Most of what I build sits on the server side: schedulers, workers, queues, databases and the APIs in front of them. I like problems where the obvious implementation is the one that quietly loses data at 3am, and the job is to notice that before a user does.

The AI work is the same discipline applied to a different component. I build retrieval pipelines and orchestration with Python, LangChain, LangGraph and FAISS, and I spend most of my effort on the parts that are not the model: whether it returns the right context, whether the output parses, and what it costs per request.

Outside of work I solve competitive programming problems and previously led research projects at Logic Play Club.

04

Experience

Jul 2026 - Present Current

Back-End AI Engineering Intern

FlyRank AI flyrank.ai

  • Architected 5 LLM powered microservices and API endpoints, integrating context aware AI into core product workflows.
  • Implemented RAG pipelines to extract and process domain context, reducing downstream generation latency by 67%.
  • Integrated the Google Gemini API with automated JSON parsing and exception handling to keep the services reliable in production.
  • Deployed automated testing and CI/CD routines, reducing deployment rollback time by 23%.
Aug 2025 - Sep 2025

AI/ML Summer Intern

Edunet Foundation edunetfoundation.org

  • Developed an end-to-end Python ML pipeline processing 7,900 spatial and temporal features to assess carbon storage sites.
  • Executed preprocessing and trained classification models on multi-variable datasets spanning 802 districts.
  • Tuned hyperparameters with scikit-learn, improving predictive accuracy from 56% to 78%.
Oct 2024 - Dec 2024

Software Engineering Intern

MIT Square mitsquare.com

  • Engineered an interactive daily quiz engine for a technical news platform, driving a 20% lift in user engagement.
  • Optimised platform-wide technical SEO and content layouts, improving organic search visibility.
  • Streamlined release cycles and code review through structured Git branch management workflows.

Unpopular Opinions

05

Selected Work

01

Distributed Job Scheduler Live

A production style task scheduling engine built around the idea that the database is the source of truth and the queue is only a notification system. Anything the queue loses can be rebuilt by reading the database again.

  • Handles 20 concurrent workers with asynchronous worker fleets, cron schedules, exponential backoff and a Dead Letter Queue.
  • Prevents race conditions and duplicate execution using atomic job claiming with FOR UPDATE SKIP LOCKED.
  • Automated dead worker recovery reached a 99.9% completion rate.
  • Node.js
  • TypeScript
  • PostgreSQL
  • Prisma
  • Docker
  • Redis
Live demo Source
02

Answer-Map AI Live

An automated assessment platform that reads handwritten answer sheets. Gemini Vision extracts questions and answers, custom bounding box logic localises each response, and a review UI puts the grading decision in front of a human.

  • Reached 78% accuracy across 25+ scanned exam papers.
  • Built custom bounding box answer localisation, a side-by-side grading UI and interactive remedial AI assistance.
  • Engineered edge case handlers for out-of-order and unattempted questions to keep evaluation stable on messy scans.
  • Gemini 2.5 Flash
  • Express.js
  • React
  • TypeScript
Live demo Source
03

Requirements-to-Test-Case Generator Colab

A retrieval tool that reads SRS documents and OpenAPI specs, then writes the test suite. Embeddings are stored in FAISS and SQLite so retrieval stays low latency, and every generated test traces back to its source.

  • Cut manual test authoring time by 60 to 70% on a real requirements document.
  • Vectorised source code with Sentence Transformers, storing embeddings in FAISS and SQLite.
  • Formulated LLM workflows to generate bidirectional, traceable unit, integration and security tests.
  • Python
  • FastAPI
  • LangChain
  • FAISS
  • Sentence Transformers
Run in Colab Source
04

ReachInbox Live Private repo

A bulk email scheduling engine with durable delivery guarantees. A campaign is written to PostgreSQL in one transaction, then each recipient becomes its own retryable job. A crash never loses work and never resends.

  • Transactional outbox pattern keeps side effects such as search indexing and Slack alerts from being lost.
  • Atomic claim with lease expiry and startup reconciliation every 60 seconds, so restarts recover cleanly.
  • Atomic Redis Lua rate limiting on whole-hour windows, with the slot returned when a send fails.
  • 75 unit and integration tests, 29 Playwright browser tests, plus failure injection and outage suites.
  • Express
  • TypeScript
  • PostgreSQL
  • Drizzle
  • BullMQ
  • Redis
  • Next.js
  • Elasticsearch
Live demo

Also built

Other work

06

Toolkit

Languages

  • Python
  • TypeScript
  • JavaScript
  • SQL
  • C++
  • Java
  • C

AI and ML

  • RAG pipelines
  • LLMs and SLMs
  • Agentic AI
  • Agent orchestration
  • LangGraph
  • LangChain
  • Gemini API
  • FAISS
  • Sentence Transformers
  • scikit-learn
  • NLP

Backend and data

  • Node.js
  • Express.js
  • FastAPI
  • Spring Boot
  • PostgreSQL
  • MySQL
  • Redis
  • BullMQ
  • Prisma
  • Drizzle ORM
  • REST APIs
  • JWT
  • Flyway

Infrastructure

  • Docker
  • Kubernetes
  • Azure
  • Linux
  • Git
  • CI/CD
  • OpenAPI
  • Playwright
  • Streamlit
  • Gradio

Systems and networking

  • Boost.Asio
  • TCP sockets
  • Async I/O
  • Thread pools
  • CMake
  • Ryu
  • Mininet
07

Education and record

Bachelor of Technology, Computer Science and Engineering

SRM Institute of Science and Technology, Chennai · 2023 to 2027

CGPA 8.58 / 10

Certifications

  • Intro to Machine Learning, NPTEL and IIT Madras2024
  • Data Science Foundations, Google2024
  • Machine Learning Using Python, INI2024
  • Cloud Computing, NPTEL2025

Beyond the coursework

  • 01 1,000+ algorithmic problems solved across LeetCode and CodeChef. 2 Star on CodeChef.
  • 02 R&D Head at Logic Play Club, directed student research groups and technical project execution.
  • 04 Shipping habit: most side projects here are deployed, not just committed.
08

Reach out to collaborate

Always open to working on something that tests my potential to the max. Email is the fastest way to reach me. I answer anything about backend systems, LLM infrastructure or distributed work.

Email mdfaiha.official@gmail.com GitHub github.com/Ziyaxoxo LinkedIn /in/muhammed-faiha Phone +91 73387 47733