Applied AI & Full-Stack Engineer
VIT Vellore · India
Cyril Sabu George. I build AI systems people can use.
From retrieval and computer vision to the APIs and interfaces around them.
01 / THE PERSON BEHIND THE CODE
Hello, I’m Cyril ↗ Applied AI
Applied AI Software Data
AI/ML experience FreshToHome
AI/ML experience GlobalLogic
VIT Vellore · Integrated M.Tech 9.39 / 10 CGPA
SELECTED WORK
Behind every output, a thoughtful system. Explore the problem, follow the workflow, then get into the engineering.
01 / Agentic AI
Time Agent From site reports to reviewed progress.
A construction report becomes an evidence-linked proposal. A person checks the match before the official progress record changes.
Evidence stays attached. People control the record.
LangGraph · FastAPI · PostgreSQL · React
Time Agent Illustrative workflow
01 Report ↗ 02 Evidence ↗ 03 Review ↗ 04 Ledger ↗ Source: sample site report
“Level 2: installed 12 window frames today.”
A supervisor reports completed work.
Extracted observation
12 window frames · Level 2
Keep the reported quantity attached to its source.
Pending approval — ledger unchanged
Activity: Level 2 window installation
A reviewer checks the proposed schedule match.
Illustrated approved update
+12 frames · source retained
Only an approved proposal becomes recorded progress.
Step 1 of 4 · Report
← Next step →
02 / Web3 · Full-stack
InvoChain An invoice with a traceable payment path.
Connect wallet verification, Ethereum invoice records, generated PDFs and IPFS document retrieval in one application.
PDF + IPFS Document generation and retrieval
React · Express · Solidity · Ethers.js
InvoChain Illustrative workflow
01 Identity ↗ 02 Invoice ↗ 03 Document ↗ 04 Payment ↗ Firebase identity + wallet verification
Account login → nonce → signature
Connect the account and prove wallet control.
Synthetic example, no transaction
Invoice #001 · 0.05 ETH
Create an invoice with recipient and amount.
Document storage path
Invoice PDF → Pinata → content ID
Generate a PDF and store it through IPFS.
Illustrative payment record
Ethereum transaction → invoice status
Confirm payment through the connected wallet.
Step 1 of 4 · Identity
← Next step →
03 / Computer vision
Fridge Vision A fridge photo. A dinner idea.
Detect ingredients in a photo, read package labels when needed, and turn structured detections into recipe suggestions.
67 Configured food classes
YOLO11 · FastAPI · OpenCV · React
Fridge Vision Illustrative workflow
01 Photo ↗ 02 Detect ↗ 03 Match ↗ 04 Cook ↗ Illustrative image contents
Tomato · Egg · Cheese
Upload a fridge image to the application.
Example detection labels
[ Tomato ] [ Egg ] [ Cheese ]
YOLO identifies ingredients; OCR can read labels.
Structured ingredient list
tomato + egg + cheese
Pass detected ingredients to the recipe matcher.
Illustrative recipe recommendation
Tomato & cheese omelette
Browse a matching recipe suggestion.
Step 1 of 4 · Photo
← Next step →
04 / Retrieval · Voice
Voice-Based RAG Ask your documents. Hear the answer.
A configurable document question-answering pipeline with text or voice input, hybrid retrieval, reranking and speech output.
3 Documented vector-store options
Python · Whisper · Chroma · Ollama · Gradio
Voice-Based RAG Illustrative workflow
01 Ingest ↗ 02 Ask ↗ 03 Retrieve ↗ 04 Answer ↗ Multi-format document ingestion
Handbook.pdf → searchable passages
Parse and chunk source documents.
Example query
“How do I submit an expense?”
Accept a typed or transcribed question.
Context selected from the documents
Dense + BM25 → ranked passages
Combine retrieval and reranking.
Illustrative pipeline, no microphone used
Retrieved context → answer → audio
Generate a response with optional speech.
Step 1 of 4 · Ingest
← Next step →
EXPERIENCE
Built with real constraints.
From retail forecasting to multimodal retrieval, I’ve worked across data pipelines, backend systems, and the details that make them reliable.
May – Jun 2026
Bangalore
FreshToHome
Machine Learning Intern
Transformed 1M+ records into model-ready datasets for perishable inventory planning. Built 3+ Airflow pipelines, added data quality checks, and automated benchmarking across 5+ forecasting models.
Airflow / PySpark / LightGBM / XGBoost / Chronos-2
Explore the workflow + Prepare Transform retail records into model-ready datasets with PySpark.
Validate Orchestrate Airflow pipelines with data quality checks.
Compare Benchmark forecasting models for perishable inventory planning.
Mar – May 2026
Noida
GlobalLogic
AI/ML Intern
Implemented modular document processing and hybrid retrieval pipelines with reranking and caching. Collaborated with a principal architect on reusable backend components and voice, transcription, and text-to-speech workflows.
Multimodal RAG / Hybrid search / Reranking / Voice interfaces
Explore the workflow + Process Turn source documents into searchable passages.
Retrieve Combine hybrid search, reranking, and caching.
Connect Support voice, transcription, and text-to-speech workflows.
Explore my related RAG project ↗
I build across data, models, and software.
I’m a Computer Science student at VIT Vellore, interested in making AI systems reliable and useful in everyday workflows.
EDUCATION VIT Vellore Integrated M.Tech · CSE 2023–2028 · CGPA 9.39/10
BEYOND THE CODE External Affairs Head, ADG VIT. Industry partnerships and outreach for AI/ML workshops, iOS Fusion 8.0, and SharkTech.
LEARNING & RECOGNITION
Oracle Cloud Infrastructure
2025 Generative AI Professional Stanford & DeepLearning.AI
Machine Learning Specialization Academic achievement
Branch Rank 6 · Semester 6 Top 15 for three consecutive years