Generative AI & RAG
LangChain retrieval pipelines, semantic caching, prompt design and FastAPI serving, taken from prototype to production traffic.
Available for SDE & AI Internships
Kaiser Mohiuddin
AI/ML Engineer · Generative AI Developer · Founder of SiratSync
Final-year B.Tech CSE student at BGSBU, Rajouri. I build and ship AI systems that reach real users — retrieval pipelines, document-forensics models and a cross-platform Islamic utility app now used by hundreds of people across India.
0+
Production systems shipped
0+
Organic Play Store installs
0%
RAG latency cut on SiratSync
0+
Chunks indexed in the retrieval store
About
Engineering that answers to people, not just to benchmarks
Assalamu Alaikum. I'm Kaiser Mohiuddin, born in Kupwara, Jammu & Kashmir, and currently in my 7th semester of B.Tech Computer Science Engineering at Baba Ghulam Shah Badshah University, Rajouri — batch 2023–2027.
My work sits at the intersection of applied machine learning and product engineering. I care less about models that look good in a notebook and more about systems people actually depend on: a RAG assistant that answers in under two seconds, a forensic pipeline that flags a forged document and can explain why, an app that keeps working when the network drops.
That instinct comes from where I grew up. Technology in the valley has to be resilient and worth the trust placed in it. For me it is an amanah — something held in trust — so I try to build things that are transparent about what they know, honest about what they don't, and useful to the person on the other side of the screen.
Outside of code I play chess and cricket, spend time with the Qur'an, and mentor juniors at university on getting their first ML project past the tutorial stage.
Computer vision
OCR, YOLOv8 detection, error-level analysis and CNN-based tampering detection for document and vehicle screening.
Shipping to real users
A published Flutter app with offline-first sync, push delivery and a live Play Store listing I maintain myself.
Leading a team
Team leader at HackHive for Smart India Hackathon 2026, coordinating a five-module build against a Ministry of Home Affairs problem statement.
Work experience & internships
Where I've built things
Founding a product, research at an NIT, and supervised-learning work in industry.
Founder & Product Owner
- Shipped a cross-platform Islamic utility app to 480+ organic Play Store installs, owning product, engineering and the store listing end to end.
- Engineered a 5,000+ chunk LangChain / FastAPI RAG pipeline backing the in-app assistant, grounded in Qur'an and Hadith sources.
- Cut response latency by roughly 40% — about 3.2s down to 1.9s — through multi-layer semantic and response caching.
- Integrated the WhatsApp Cloud API so users can reach the assistant without installing anything.
FlutterFirebaseLangChain
FastAPIWhatsApp Cloud APIPlay Console
AI/ML Research Intern
National Institute of Technology, Srinagar
- Researched classical ML pipelines under Prof. Hashmat, covering feature engineering and evaluation methodology.
- Built preprocessing and normalisation for Kashmiri Roman-transliterated text, a low-resource setting with no standard orthography.
- Trained and evaluated sentiment classifiers using TF-IDF features with a Linear SVM baseline.
PythonScikit-learnTF-IDF
Linear SVMNLP
AI/ML Intern
Winnovation
- Trained and evaluated supervised models across classification and regression tasks.
- Delivered a diabetes detection capstone, covering data cleaning, model selection and metric reporting.
- Built a vehicle license plate recognition system combining detection with OCR.
PythonScikit-learnPandas
OpenCVTesseract OCR
Flagship product
SiratSync — consistency, not streak anxiety
An Islamic habit companion that helps people stay steady in Salah, Qur'an and daily good deeds. It keeps prayer times, Qibla, tasbih, duas and the 99 Names in one place, works offline, and now answers questions through a retrieval-grounded assistant.
- Salah and Qur'an tracking that rewards returning, not perfection
- Offline-first architecture with Firestore sync on reconnect
- Assistant grounded in a 5,000+ chunk Qur'an and Hadith index
- Quiet, faith-aligned reminders through Firebase Cloud Messaging
- Progress views that show consistency over weeks, not days
Selected projects
Systems I've designed, built and deployed
Every link below goes to live code, a live deployment or a recorded demo.
SiratSync — Mobile Application
A cross-platform Islamic utility app with habit tracking, prayer times, Qibla, tasbih, duas and the 99 Names. Built offline-first so the core features keep working without a connection, with Firestore reconciling state once the device is back online. Published and maintained on the Play Store with 480+ organic installs.
FlutterDartFirebase Auth
Firestore SyncFCMOffline-First
Sirat Assistant — Conversational Islamic AI
A retrieval-grounded assistant serving SiratSync users. Queries are routed by intent, answered from a 5,000+ chunk Qur'an and Hadith index rather than model memory, and served through FastAPI. Semantic caching brought median response time down from roughly 3.2 seconds to 1.9 seconds; a WhatsApp Cloud API channel makes it reachable without the app.
PythonLangChainLLaMA
RAGFastAPISemantic Caching
WhatsApp Cloud API
Smart India Hackathon 2026
VeriShield AI
Team Leader @ HackHive · PS 26188 (MHA / SSB)
An AI-based fake identity and document screening system for the Ministry of Home Affairs / SSB problem statement. We are actively working on this project with an MVP ready currently — implementing document OCR text extraction, layout forensics, and digital tampering detection, with additional verification modules in progress.
PythonOpenCVFastAPI
PaddleOCRDeep LearningDocker
AutoStream AI Agent
A stateful lead-generation agent for a SaaS video platform, built on LangGraph. It detects intent, retrieves product facts through RAG rather than improvising them, and collects contact details across a multi-turn conversation — firing actions only when the conversation state says it should. Containerised with Docker Compose.
PythonLangGraphRAG
FastAPIDocker Compose
Kashmiri Sentiment Analysis
Sentiment classification for Roman-transliterated Kashmiri, developed during the NIT Srinagar research internship. The hard part is the input: no standard spelling, heavy code-mixing with Urdu and English, and almost no labelled data. The pipeline handles normalisation and TF-IDF featurisation, with a Linear SVM baseline served over Flask.
PythonScikit-learnTF-IDF
Linear SVMFlask API
License Plate Recognition & Blacklist Verification
Real-time ANPR built on a custom-trained YOLOv8 detector with Tesseract OCR for plate reading. Recognised plates are checked against a watchlist using fuzzy matching, so a single misread character doesn't cause a miss — the practical failure mode in field conditions with dirt, glare and motion blur.
PythonYOLOv8 (custom-trained)OpenCV
Tesseract OCRFuzzy Matching
Task Manager Pro
A production-shaped task system: JWT auth with hashed credentials, role-based access for admin and standard users, full CRUD with filtering and search, a statistics dashboard, rate limiting and request validation. Documented with Swagger and deployed as a containerised stack that comes up with a single Docker Compose command.
React 18Node.jsExpress
MongoDBDocker ComposeJWT
RBACSwagger
Library Management System
A deployed library system with separate admin and student portals: authentication with auto-generated temporary passwords and forced first-login reset, book and member management, issue and return tracking, and automatic fine calculation. Written in PHP against MySQL using PDO prepared statements throughout.
PHP (PDO)MySQLHTML5CSS3
Other sites I've built and shipped
Client and community front-ends, each live and maintained.
Technical stack
What I reach for
Tools I've used to ship something real, not just to complete a tutorial.
Languages
PythonDartJavaScript
CC++PHP
SQLHTML5 / CSS3
AI / Machine learning
Scikit-learnTensorFlowKeras
NumPyPandasTF-IDF
Linear SVMCNNs
Generative AI
LangChainLangGraphRAG
LLaMAGroq APIVector stores
Semantic cachingPrompt engineering
Computer vision
OpenCVYOLOv8PaddleOCR
Tesseract OCRError-level analysis
Face embeddings
Backend & data
FastAPIFlaskNode.js
ExpressMongoDBMySQL
FirestoreREST APIs
Mobile & frontend
FlutterReact 18Vite
Firebase AuthFCMOffline-first sync
Tooling & deployment
DockerDocker ComposeGit / GitHub
SwaggerPlay ConsoleLinux
Jupyter
Working style
Team leadershipProduct ownership
Technical writingMentoring
Problem decomposition
Education & credentials
Academic background
B.Tech, Computer Science Engineering
2023 — 2027Baba Ghulam Shah Badshah University (BGSBU)
Rajouri, Jammu & Kashmir, India
Final-year student, currently in the 7th semester. Coursework concentrated in AI/ML and systems, alongside independent product work and hackathon engineering.
Artificial IntelligenceMachine Learning
Data Structures & AlgorithmsDesign & Analysis of Algorithms
Database Management SystemsOperating Systems
Computer NetworksSoftware Engineering
Computer Organization & ArchitectureTheory of Automata
Object-Oriented ProgrammingWeb Technologies
Higher Secondary (12th), Science — PCM
2020 — 2021Government Higher Secondary School, Drugmulla
Kupwara, Jammu & Kashmir, India
Completed with distinction. Built the mathematics and physics grounding that later made the machine-learning work approachable.
PhysicsChemistryMathematics
Informatics PracticesEnglish
Contact
Open to SDE and AI internship roles
Hiring, collaborating, or just want to talk about retrieval systems? Write to me.
Based in
Kupwara, Jammu & Kashmir, India
Open to remote, hybrid and relocation