Gaurav
visibility ···
Gaurav Singh
location_onVellore, India

Gaurav Singh

ML Engineer & AI Researcher Full-Stack Architect Open Source Developer
codeGitHub workLinkedIn devicesGoogle Dev terminalLeetCode

About

  • Final-year CSE @ VIT Vellore · 8.11 CGPA
  • Specialising in AI / ML, Cloud Computing & System Design
  • Building scalable solutions for healthcare & environmental intelligence
  • Former Technical Head — GeoSpatial Club, leading 15+ developers

Certifications

verified
Google Cloud ACE
Google Associate Cloud Engineer · Google Cloud · 2026
smart_toy
Agentic AI
IBM · 2026
psychology
AWS AI Practitioner
AIF-C01 · AWS · 2026
cloud
OCI AI Associate
Oracle Cloud Infra 2025 AI Foundations · Oracle · 2025
search
Ethical Hacking
Footprinting & Reconnaissance · LinkedIn Learning · 2024

Projects

Bio-SentinelX
Capstone Research · Dec 2025 – Mar 2026
Winner at NIT Durgapur Hackathon
  • •Engineered multi-provider LLM orchestration middleware (Gemini 2.0, Groq Llama 3, Claude Sonnet, OpenRouter, Ollama) with CoT reasoning, automatic provider failover, and LRU prompt caching — reducing end-to-end inference latency by 40%.
  • •Architected dual-embedding RAG pipeline (Supabase pgvector, hybrid dense + sparse retrieval, TF-IDF chunking, 3-tier context memory) with real-time retrieval across OpenAlex (200M+ works), PubMed, WHO/CDC, ClinicalTrials.gov.
  • •Deployed stacked ensemble (XGBoost + RF + NN, 98.37% accuracy, 11 classes) with in-browser TFLite quantised inference; containerised via Docker + Kafka, shipping 155+ zero-downtime releases via GitHub Actions CI/CD.
React 19 FastAPI Supabase TFLite
Bio-SentinelX
Bio-SentinelX
Arogya AI
Research Project · Aug 2025 – Nov 2025
  • •Designed a dual-engine AI architecture fusing a Random Forest classifier (99%+ accuracy, 819 features, SMOTE-balanced 4,201–20,748 samples) with Gemini 2.5 Pro for contextualised, multi-label prediction across 399 ICD-mapped diseases.
  • •Developed an 807-dim TF-IDF NLP vectorisation pipeline with high-precision cosine similarity matching, incorporating XAI/SHAP explainability to provide transparent insights into complex medical model predictions.
  • •Engineered secure RBAC practitioner and patient portals, implementing clinical safety guardrails that automatically mask low-confidence predictions to ensure patient data integrity.
Scikit-learn Gemini 2.5 Pro SHAP SMOTE
Arogya AI
Arogya AI
ShakeAlert
CDMM VIT Internship · May 2025 – Jul 2025
  • •Built a full-stack earthquake early warning system for India — low-latency seismic telemetry parsing with real-time Azure Maps coordinate dispatch and alerting.
  • •Integrated Firebase Firestore streaming pipelines with Pandas-based signal anomaly identification against live seismic sensor feeds from India's NDMA network.
  • •Implemented Python Seismology APIs to classify P-wave and S-wave arrivals, enabling sub-second alert dispatch to potentially affected regions.
Firebase Azure Maps Pandas Python
ShakeAlert
ShakeAlert

Experience

Technical Head
Mar 2025 — Jan 2026
GeoSpatial Club · VIT Vellore

Supervised technical development for VIT's mapping society with 100+ members. Led and mentored a team of 15+ developers, delivered workshops in Remote Sensing, and orchestrated MapQuest – The GIS Treasure Hunt (Sept 2025) as the club's flagship event.

Software Development Intern
May 2025 — Jul 2025
CDMM · VIT Vellore

Developed ShakeAlert — a full-stack, scalable earthquake early warning system for India. Built telemetry parsing pipelines and Pandas-based signal anomaly identification from live seismic feeds.

Open Source Contributor
Active
roadmap.sh

Contributing to community-driven educational resources for software engineering. Developed and refined technical roadmaps for ML and Full-Stack domains.

Deep learning framework — used in Bio-SentinelX ensemble & Arogya AI DenseNet training. Retrieval-Augmented Generation — powers Bio-SentinelX's dual-embedding medical search across 200M+ papers. LLM orchestration library — used for CoT reasoning chains and provider failover logic. Parameter-Efficient Fine-Tuning — enables fine-tuning large models with minimal compute. TensorFlow Lite — runs quantised Bio-SentinelX model directly inside the browser. Facebook AI Similarity Search — used for blazing-fast approximate nearest-neighbor retrieval.

Skills

psychology ML / AI / LLM
PyTorch TensorFlow/Keras Scikit-learn XGBoost SMOTE SHAP LangChain LlamaIndex Hugging Face RAG Prompt Engineering LoRA/PEFT Agentic AI TFLite Supabase pgvector FAISS BERT Transformers Hybrid Retrieval
cloud_sync MLOps / Cloud / Stack
Docker Kafka Redis FastAPI GitHub Actions CI/CD AWS (EC2, S3, Lambda) GCP Railway Vercel
terminal Field of Interest
OS Linux Ubuntu Debian Arch Linux Kali Linux WSL PowerShell Sh Git Bash Windows Terminal Command Prompt Bash
database Languages & Databases
SQLite DynamoDB Firebase MySQL Netlify Google Cloud Python TypeScript/JS Java C/C++ SQL Bash FrontEnd HTML CSS Javascript Backend
groups Professional Skills
Problem Solving Cross-functional Collaboration Agile Methodology Technical Mentorship

Distinctions & Awards

IIC NIT Durgapur Ideathon 2026
Bio-SentinelX — Disease prediction & analysis platform
1st Place · 9.5/10May 2026
Anveshana 3.0 National Challenge
BMSIT&M, Bangalore · Top 50 nationwide
Grand FinalistMar 2026
India Innovates 2026
Municipal Corp of Delhi · Bharat Mandapam, New Delhi
Grand FinalistMar 2026
ET Gen AI Hackathon
Economic Times · Top Teams Nationwide
Phase 2 QualifierMar 2026
Amazon ML Summer School 2026
Amazon Science · Competitive ML Selection Programme
Top 2.2% · Cohort2026

GitHub Activity

Contribution Calendar · github.com/gaur-avvv
View Profile →
GitHub contribution chart for gaur-avvv
GitHub streak stats
GitHub stats
Contact
Vellore Cable Station · Terminal 04
sensors
Despatch Nodes

Open to engineering roles · July 2026 · Vellore, India.