████████████████████████████████████████████████████████████  ██╗  ██╗███████╗██╗     ██╗      ██████╗
████████████████████████████████████████████████████████████  ██║  ██║██╔════╝██║     ██║     ██╔═══██╗
███████████████████████████████████`.        ╙██████████████  ███████║█████╗  ██║     ██║     ██║   ██║
████████████████████████████████▀  ¿▓▓▓▓▓▓▓▓▄/ "████████████  ██╔══██║██╔══╝  ██║     ██║     ██║   ██║
██████████████████████████████▀.  ▓▓▓▓▓▓▓▓▓▓▓▓   ▐██████████  ██║  ██║███████╗███████╗███████╗╚██████╔╝▄█╗
██████████████████████████████ `  ▓▓▓▓▓▓▓▓▓▓▓▓  ` ██████████  ╚═╝  ╚═╝╚══════╝╚══════╝╚══════╝ ╚═════╝ ╚═╝
██████████████████████████████ `  ▓▓▓▓▓▓▓▓▓▓▓▓   ▄██████████
▀██████████████████████████████▌  ▀▀▓▓▓▓▓▓▓▌╓╖. ████████████  ███╗   ██╗██╗ ██████╗███████╗  ████████╗ ██████╗
█▄▀██████████████████████████████▄ ╩╦╙▀▀▀▀▀ ╣`,█████████████  ████╗  ██║██║██╔════╝██╔════╝  ╚══██╔══╝██╔═══██╗
▄▀█▄╙█████████████████████▀▀▀▀█████▄▄ .... ,▄███████▀███████  ██╔██╗ ██║██║██║     █████╗       ██║   ██║   ██║
██▄▀█▄╙█████████████████▀  ╪╢%╦══~╓,└ ╚▒▒▒ ╙▀|,╓╓═╤H   ▀████  ██║╚██╗██║██║██║     ██╔══╝       ██║   ██║   ██║
█▀▀▀-▀█▌▄▀█████████████   ║▒▒▒▒▒▒▒▒▒▒╢╦ ╘ -╣▒▒▒▒▒▒▒▒▒╢╕   ▀█  ██║ ╚████║██║╚██████╗███████╗     ██║   ╚██████╔╝
██▄▀██└║▄▄▄████████████▄          ═╕╕╕╕╕═╕═══════       ▄▄▄▄  ╚═╝  ╚═══╝╚═╝ ╚═════╝╚══════╝     ╚═╝    ╚═════╝
████▄▀█▌║███  ████████▌         ╕   ╩▒▒▒▒▒▒▒▒▒Ñ          ███
██████▌Ö▓▌   ▀██████████`╔▒▒╣ █ ▒▒m   ╚▒╢▒▒▒╩ -╣▒ ▌ ▒▒▒ ████  ███╗   ███╗███████╗███████╗████████╗  ██╗   ██╗ ██████╗ ██╗   ██╗
████ -"" ∞╙,▀.╙▀███████╜ ▒▒▒ ▄█ Ñ   -   S.  ═▒▒▒▒ █ ║▒▒╕└███  ████╗ ████║██╔════╝██╔════╝╚══██╔══╝  ╚██╗ ██╔╝██╔═══██╗██║   ██║
████████▄ -«   ∞▄.▀",╓═     ╒██   ═╣▒▒ `Ñ╛        █▌ ▒▒▒ ███  ██╔████╔██║█████╗  █████╗     ██║      ╚████╔╝ ██║   ██║██║   ██║
█████████▌ º     ╤╣▒╣╩^",▄▄███▀  ▒▒╣"     ''''''' ▀▀     `██  ██║╚██╔╝██║██╔══╝  ██╔══╝     ██║       ╚██╔╝  ██║   ██║██║   ██║
█████████  ▌       ▄▄████████─         ---------    L'▒▒▒ ██  ██║ ╚═╝ ██║███████╗███████╗   ██║        ██║   ╚██████╔╝╚██████╔╝
▀▀▀▀▀▀▀▀▀▀▀▀▀-     ▀▀▀▀▀▀▀▀▀▀       '╧╧╧╧╧╧╧╧╧`     ╚ ╧╧╧- ▀  ╚═╝     ╚═╝╚══════╝╚══════╝   ╚═╝        ╚═╝    ╚═════╝  ╚═════╝

AI Systems Builder • Agentic AI • LLMs • Multimodal Intelligence • Edge AI

I design, build, and deploy production-grade AI systems that combine LLMs, agentic reasoning, retrieval pipelines, multimodal perception, and efficient deployment. My work sits at the intersection of research depth and engineering rigor.

I’m an AI/ML Engineer and system builder focused on transforming cutting-edge research into robust, scalable, real-world AI systems.

My experience spans Generative AI, agentic workflows, RAG architectures, multimodal pipelines, computer vision, and edge optimization, with strong emphasis on clean system design, reproducibility, and performance under constraints.

I enjoy breaking complex problems into modular, testable, production-ready AI pipelines—from data ingestion to deployment.

• 🤖 Autonomous Agentic Systems (tool-augmented, multi-agent, reasoning-driven workflows)
• 🧠 Advanced RAG Architectures for long documents, scientific data, and domain-heavy knowledge
• 🎙️ Voice-based AI Systems (speech → reasoning → action pipelines)
• ⚙️ Efficient AI for Edge & Resource-Constrained Devices
• 🌐 End-to-end AI products integrating cloud, APIs, and real-time inference

• 🤖 Agentic AI Systems — multi-tool orchestration, research agents, decision pipelines
• 📚 Retrieval-Augmented Generation (RAG) — embeddings, vector stores, structured retrieval
• 👁️ Vision & Multimodal Pipelines — VLMs, OCR-free document intelligence, scene understanding
• 📱 Edge & Embedded AI — INT8 quantization, pruning, Raspberry Pi & Jetson deployments
• 🌐 Full-Stack AI Applications — APIs, dashboards, cloud-native pipelines
• 🐳 MLOps & Deployment — Dockerized, reproducible, scalable AI workflows

• Large Language Models (LLMs), Transformers
• Retrieval-Augmented Generation (RAG), Embeddings
• Multimodal AI, Vision-Language Models
• Prompt Engineering, Structured Reasoning

• LangChain, n8n
• Tool-Augmented LLMs
• Multi-Agent Workflows

• INT8 Quantization, Structured Pruning
• Efficient Inference
• Resource-Constrained Deployment

• Docker, GCP, Firebase
• FastAPI, Flask, REST APIs

• SQL, NoSQL, MongoDB
• Pinecone, Firebase Realtime DB

• Smart Contracts
• Token-Based Systems
• Digital Signatures

• Operating Systems
• DBMS, Computer Networks
• System Design, DSA

I’ve worked on many applied AI systems across domains—below are representative projects showcasing depth, scale, and system design.

LangChain • LLMs • HuggingFace • Web APIs

• Designed an agentic research assistant orchestrating multiple LLM-powered tools
• Implemented agents for literature retrieval, summarization, and structured insight generation
• Integrated academic sources (ArXiv) and web search APIs
• Reduced manual literature review by auto-generating consolidated research notes

n8n • Pinecone • Gemini Embeddings • Docker

• Built a full RAG pipeline to extract structured medical insights from unstructured PDFs
• Automated ingestion, chunking, embedding, retrieval, and summarization using n8n
• Enabled contextual Q&A over medical documents
• Fully containerized for reproducible experimentation

Donut • BART • FLAN-T5 • INT8 Quantization

• Designed a modular multimodal pipeline for long, layout-complex scientific PDFs
• OCR-free layout understanding using Donut
• Fine-tuned summarization and QA models on large scholarly datasets
• Applied INT8 quantization across pipeline
• 📈 Results:
  • ~70% model size reduction
  • ~40% faster inference
  • +15% ROUGE-L, +28% METEOR
  • 85.3% QA accuracy
• Conducted ablation studies for system analysis

YOLOv12 • Raspberry Pi • Jetson Nano

• Optimized object detection models using pruning + INT8 quantization
• Reduced model size by 65% with accuracy preservation
• Designed an embedded ADAS pipeline integrating:
  • Object detection
  • Lane detection
  • Ultrasonic sensing
  • GPIO-based control
• Achieved ~210 ms end-to-end latency under CPU-only constraints

LangChain • Gemini APIs • yFinance

• Built a multi-agent financial analysis system
• Implemented data analysis, market reasoning, and decision-support agents
• Generated natural language insights from time-series data
• Designed for automated decision-support simulation

BLIP • HuggingFace

• Implemented a vision-language pipeline for structured scene descriptions
• Fine-tuned BLIP on custom datasets
• Integrated detection, tracking, and counting for real-time context-aware captions
• Enabled downstream reasoning by LLMs

• NFC / QR Token Payment Prototype — secure peer-to-peer transfers
• Blockchain Certificate Issuance Platform — tamper-resistant verification via smart contracts