Fresh B.Tech graduate in AI & Data Science β I build automation systems that solve real problems. Not just prototypes β I've shipped workflows that handled 3200+ emails without hitting Gmail limits, automated an entire hackathon backend end-to-end, and built AI pipelines that generate and sell stock images passively. I think in systems, not scripts.
| Domain | What I've shipped |
|---|---|
| π€ AI Agents | Local RAG agent (private LLM + document retrieval) |
| β‘ n8n Automation | 10+ production workflows across events, content, fraud & outreach |
| π§ LLM Pipelines | Gemini + OpenAI prompt chains for image metadata, fraud detection |
| π ML & Data | Models & datasets β actively building |
Free-tier first. Every workflow above runs on free API limits. I solve rate limits with batched loops, staggered waits, and parallel chains β not by throwing money at the problem.
- Error monitoring across all workflows β instant Telegram + Gmail alerts
- Google Sheets as a lightweight ops dashboard (status, logs, deduplication)
- Modular design β workflows are reusable and easy to hand off
- π Building a Local RAG Agent β private LLM + document retrieval, fully offline
- π Planning AI agent pipelines, ML models & LLM fine-tuning projects
- π€ Open to AI automation freelance, internships & collaborations
n8n Python LangChain Ollama Google Gemini OpenAI HuggingFace
Raspberry Pi Arduino FAISS ChromaDB Google Workspace APIs Telegram Bot API
Built the entire backend automation for a company-organized hackathon.
- AI reads payment screenshots β extracts UPI transaction ID, amount, bank name
- Duplicate UPI detection to prevent payment fraud
- Automated email pipeline: welcome β reminder β event day β winner announcement
- Sent to HOD, Principal, company & participants β zero manual effort on event day
- Stack: n8n Β· OpenAI Vision Β· Gmail Β· Google Sheets
Built a rate-limit-aware email system for my startup's data collection survey.
- Engineered send pacing + wait logic to stay within Gmail's daily limits
- Dual-loop system: survey email β 2hr gap β thank you email on response
- Telegram notifications for real-time send status monitoring
- Stack: n8n Β· Gmail Β· Google Sheets Β· Telegram
End-to-end automation to generate, enhance, and list AI stock images for sale.
- Bulk prompt engineering via 10 parallel Gemini LLM chains (prompt β 10 variations)
- 15 parallel HuggingFace model calls for bulk image generation
- Gemini Vision auto-generates SEO metadata (title, tags, description) per image
- Rate-limit management: batched loops + timed waits across all free-tier APIs
- Stack: n8n Β· Google Gemini Β· HuggingFace Β· Google Drive Β· Google Sheets
Real-time scam detection tool with Telegram + web inputs.
- Accepts text or image input via Telegram bot or website webhook
- LLM classifies message as SCAM / LEGIT / UNKNOWN with confidence score + reason
- Built the full backend; team handled data collection and frontend
- Stack: n8n Β· OpenAI Vision Β· Google Gemini Β· Telegram Β· Webhook
Fully automated video creation and upload pipeline β no human in the loop.
- Reads content ideas from sheet β AI generates script & caption
- Calls video generation API β polls status β fetches file β uploads to YouTube
- Error logging + status tracking back to Google Sheets
- Stack: n8n Β· OpenAI Β· YouTube API Β· Google Sheets
Auto-generates personalized certificates at scale.
- Copies Google Slides template per participant β replaces name/details β converts to PNG
- Generates unique QR code per certificate, cleans up temp files after
- Status tracking per certificate in Google Sheets
- Stack: n8n Β· Google Slides Β· Google Drive Β· QR API
"I don't just automate tasks β I automate outcomes."