Hi, I'm Simran Shaikh
🎯 Role → Generative AI & ML Engineer (Open to AI/ML · GenAI · Automation roles)
📍 Location → Pune, India (Remote-friendly)
🎓 Education → B.E. Computer Science, MSU Vadodara — CGPA 8.24
🏆 Wins → 3× Hackathon Winner (Global Agent Hackathon · Agentic Postgres · IIT Bombay)
⚡ Superpower → Building multi-agent LLM systems that ship to production, not just demos
🔬 Research → Published 12-class industrial defect dataset on Kaggle (ORCID registered)
🌍 Open Source → Top 5% of 7,000+ contributors globally — GSSoC Extended 2024
Production-grade AI systems — multi-agent orchestration, RAG pipelines, LLM integrations, and computer vision models that operate under real constraints.
| Domain | What I've Shipped |
|---|---|
| 🤖 Agentic AI | 6-agent parallel research system · 4-agent code review platform · conversational deployment agent |
| 🔍 RAG & LLM | Semantic retrieval pipelines · dynamic prompt engineering · structured output agents |
| 👁️ Computer Vision | Industrial defect detection (ResNet + EfficientNet, 90%+ acc) · autoencoder anomaly detection |
| ⚙️ Automation | n8n · Zapier · MCP · webhook-driven event pipelines · REST API chaining |
| 📊 Data | ETL pipelines · 100K+ record scraping at scale · CI/CD-scheduled workflows |
| 🏆 | Award | Details |
|---|---|---|
| 🥇 | Global Agent Hackathon — 1st Place | Agentic LLM system competing globally |
| 🥇 | Agentic PostgreSQL Challenge — 1st Place | Hosted by TigerDB & DEV Community |
| 🥈 | Bhashathon — IIT Bombay — 2nd Place | NLP & language AI competition |
| 🥇 | 1st Rank — GTU Diploma | Academic Excellence Certificate |
| 🥈 | ACPC Gujarat State Rank: 7th | GTU affiliated competitive programming |
| ⚡ | GSSoC Extended — Top 5% of 7,000+ | Pull Shark ×2 · Quickdraw · Starstruck |
| 🎓 | Gen AI Academy Certification | Google Cloud & Hack2Skill |
| 🔬 | Published Kaggle Dataset (ORCID) | 12-class casting defect detection benchmark |
🧠 MindMesh AI — 6-Agent Parallel Research System
ReactTypeScriptGemini 2.5 Flash/ProSupabase Edge Functions
Two-phase parallel LLM architecture: Phase 1 runs Research + Pro/Con Advocate agents via Gemini 2.5 Flash; Phase 2 runs Bias Checker + Fact Verifier via Gemini 2.5 Pro; Synthesizer delivers confidence-scored output with real-time streaming. 3–5s analysis vs 20s+ sequential · 🌐 Live Demo
🤖 Multi-Agent Code Review System — 🥇 Hackathon Winner
TypeScriptReactPostgreSQLTigerDBLLM Agents
4 specialised AI agents (Quality · Security · Performance · Docs) running in parallel with human-in-the-loop review gates via TigerDB zero-copy forks. 4× faster analysis (40s → 10s), zero storage overhead. 1st place — Agentic PostgreSQL Challenge
🔍 SEO InsightHub — Production GenAI Agent ⭐ 20 Stars
PythonAgnoGroq LLMFireCrawlExa APIStreamlit
Autonomous SEO audit agent that crawls sites, runs RAG-powered technical audits, keyword gap analysis, and competitor benchmarking — delivering GDPR-compliant dashboards. +20% client organic performance.🌐 Live Demo
PythonPyTorchOpenCVResNet-50EfficientNet-B3NumPy
End-to-end defect detection system: 500+ images/min · ResNet-50 + EfficientNet-B3 ensemble · MixUp/CutMix augmentation · Test-Time Augmentation · 90%+ accuracy across 12 defect classes · training data expanded 3×.
PythonLangChainFAISSGroq LLMScrapyStreamlit
Agentic pipeline: scrapes job listings → embeds portfolio into FAISS vector store → RAG + dynamic prompt engineering → hyper-personalised cold emails. +25% recruiter response rate.
ZapierWhatsApp APILLMn8n
End-to-end WhatsApp automation agent handling student queries autonomously — no human in the loop for routine support flows. FAQ resolution · escalation routing · session management.
TypeScriptn8nMixtralWebhook
n8n + Mixtral workflow for automated professional LinkedIn post generation — prompt in, polished SEO-optimised content out.
🏭 AI / ML Intern — Atlas Copco(Jan 2026 – Present) · Pune, On-site
- LLM Automation: Built a Python CLI agent chaining 10+ REST APIs (Azure AD OAuth2) using function-calling patterns, replacing a 15-step manual workflow with one command — 80% faster onboarding
- Computer Vision: Implemented ResNet transfer learning + autoencoder anomaly detection in PyTorch — 90%+ accuracy, feature dimensionality reduced 65%
- Data Pipeline: Processed 10,000+ industrial images through automated augmentation pipelines; expanded training data 3×, improved model accuracy 15% across 5 product categories
- Built 5+ event-driven scraping pipelines (Scrapy · BeautifulSoup · Pandas) processing 100,000+ records
- Docker-containerised CI/CD-scheduled workflows → 60% reduction in manual collection, 10+ hrs/week saved
- Integrated REST APIs with XPath/CSS parsing and fault-tolerant retry middleware
- Top 5% of 7,000+ contributors globally · Pull Shark ×2 · Quickdraw · Starstruck achievements
- Merged 20+ PRs across LLM tooling and agent framework repositories
- Reviewed 30+ peer PRs, mentoring contributors and reducing review cycle time 25%
📦 12-Class Casting Defect Detection Dataset — Kaggle
The only publicly available multi-class extension of the binary casting defect benchmark. Paired with a ResNet-50 + EfficientNet-B3 ensemble using MixUp/CutMix augmentation and TTA — 90%+ accuracy across all 12 defect classes. ORCID: 0009-0002-2854-8468
