GET /resume

available for backend / AI engineering roles

Vebhav
Sharma

Backend Developer · Artificial Intelligence Engineer

I design and ship the layer users never see — REST APIs, data models, and AI inference pipelines that hold up under real traffic. Currently building with FastAPI, Django and local/hosted LLMs.

View projects Get in touch

GET /about HTTP/1.1200 OK

Summary

Computer Science undergraduate specializing in Artificial Intelligence, focused on backend engineering with FastAPI, Django, and PostgreSQL / MySQL. Backend contributor across three team projects covering REST API design, OCR pipelines, and AI model integration (Hugging Face, local LLMs). Solved 300+ LeetCode problems and I'm comfortable owning the API and database layers in a team setting.

LeetCode solved
300+
Team projects shipped
03
CGPA
9.48
Grad year
2028

GET /skills HTTP/1.1200 OK

Stack & skills

response.json
"languages" :
JavaPythonC
"backend" :
FastAPIDjangoREST APIs
"ai_ml" :
Hugging FaceOllama (Local LLMs)LangChainEasyOCRNumPyPandasMatplotlibPydanticSpeech RecognitionText-to-Speech
"web" :
HTMLCSSJavaScript
"databases" :
PostgreSQLMySQLFirebase
"tools" :
GitGitHub
"cs_fundamentals" :
Data Structures & AlgorithmsOOPDBMSOperating SystemsComputer Networks

GET /projects HTTP/1.1200 OK

Selected projects

POST /projects/ai-scam-detection-calling-app shipped

AI-Powered Scam-Detection Audio Calling App

Mobile & AI Engineer · Sep 2026

  • Built a cross-platform (Android/iOS/Web/Desktop) real-time voice calling app using Flutter and WebRTC, with a custom WebSocket signaling service handling connection lifecycle, heartbeats, and auto-reconnect.
  • Integrated an AI voice-detection pipeline that records a 10-second sample of the remote caller's audio and sends it to a Hugging Face-hosted model to flag likely scam calls in real time.
  • Implemented live scam-number synchronization over WebSocket with local persistence (shared_preferences) for offline access to flagged numbers.
  • Added a premium subscription tier with Razorpay payment integration (order creation + signature verification against a deployed backend).
  • Built an FFmpeg-based audio noise-removal pipeline that cleans captured audio before AI analysis, routed through a separate deployed backend service.
  • Wrote 18 automated test files (unit + widget tests) covering the call controller, WebRTC service, signaling, scam detection, and payment flows.
FlutterDartWebRTCWebSocketHugging Face APIRazorpayFFmpeg
View code →
POST /projects/ai-revenue-recovery-agent shipped

Multi-Modal AI Revenue Recovery Agent

AI Engineer · Sep 2026

  • Engineered a schema-validated LLM classification layer with Google Gemini 1.5 Flash + Instructor, transforming unstructured payment-failure signals into 3 typed diagnostic categories without any regex-based parsing.
  • Orchestrated a 5-node stateful reasoning pipeline using LangGraph, chaining webhook ingestion, AI diagnosis, and conditional routing into one auditable decision flow.
  • Deployed 3 differentiated recovery channels driven by model output — automated retries, a LiveKit WebRTC voice call in Hinglish for transactions above ₹1000, and a Telegram payment link for lower-value failures.
  • Capped autonomous agent actions at 3 interventions per transaction and logged every classification decision to an immutable audit trail, ensuring bounded execution and full explainability.
  • Validated pipeline reliability through 6+ automated test cases covering HMAC-SHA256 signature verification, classification-to-action mapping, and idempotent transitions across a 5-stage transaction lifecycle.
PythonFastAPIGoogle GeminiLangGraphInstructorRazorpay APILiveKitTelegram Bot API
View code →
POST /projects/ai-audio-classification-api shipped

AI Audio Classification API

Backend Developer · Team Project · Feb–Mar 2026

  • Owned backend API design for a multilingual AI audio classification service supporting Tamil, English, Hindi, Malayalam and Telugu.
  • Implemented AI vs. human voice detection using Facebook Wav2Vec2 feature extraction.
  • Built endpoints returning structured JSON responses with calibrated confidence scores.
  • Designed the FastAPI service layer to keep business logic out of route handlers for maintainability.
PythonFastAPIWav2Vec2
View code →
POST /projects/bhojanam shipped

Bhojanam — Offline AI Food Label Copilot

Backend Developer · Team Project · Nov–Dec 2025

  • Built backend logic for an offline AI assistant that analyzes food nutrition labels with zero cloud dependency.
  • Developed OCR pipelines supporting images, screenshots and live camera input.
  • Integrated local LLMs into the backend for on-device, privacy-preserving inference.
OCRLocal LLMsAI
View code →
PUT /projects/1code in progress

1Code — Developer Assessment Platform

Fullstack Developer · Jul 2026

  • Built a developer assessment platform testing how developers debug and collaborate with AI tools — "Diagnose first. Prompt second."
  • Designed the backend with FastAPI and PostgreSQL, deployed on Render, with the frontend targeting Vercel.
  • Integrated the Gemini API to power AI-based code evaluation.
  • Implemented a Submissions table to persist evaluation results during development.
PythonFastAPIPostgreSQLGemini API
View code →

GET /credentials HTTP/1.1200 OK

Certifications & education

Certifications

  • FINALIST AI Summit 2026 Finalist
  • CERT Mastering MySQL
  • CERT SQL Database Certification — Oracle
  • WORKSHOP Google Development Workshop — Firebase, MediaPipe, Antigravity

Education

Galgotias University, Greater Noida

B.Tech, Computer Science & Engineering (Artificial Intelligence)

CGPA 9.48 Expected graduation 2028

Colonels Brightland Public School, Agra

CBSE Schooling

Class XII 83% (2023) Class X 80% (2021)

POST /contact HTTP/1.1

Let's build something

Open to backend and AI engineering roles, internships and collaborations. Send a request below or reach me directly.

location Greater Noida, Uttar Pradesh, IN
email
phone +91 7037627734
links
github.com/vebhav-code linkedin.com/in/vebhav-sharma
currently accepting new opportunities