AMAR SAISH

Computer Languages: Python, C++, Javascript, SQL, Julia

Frameworks / Libraries: PyTorch, TensorFlow, LangChain, LangGraph, LlamaIndex, Pydantic, CrewAI, Hugging Face, Docling, Streamlit, TypeScript, ComfyUI, Data extraction and orchestration

AIML: Machine Learning, Deep Learning, NLP, LLMs, Docker, Kubernetes, CI/CD, GANs, Transformers, RAG Systems, Multi-Agentic Workflows, Vision Models, OCR, Computer Vision, Data Scraping, Model Optimization (Quantization / Caching)

Software Tools: AWS, GCP, Azure, Cursor, GitHub, Power BI, VS Code

• Led the end-to-end forward deployment of a production-grade pipeline automating information extraction and insights from various structured / unstructured data, and data ingestion to platforms. Architected enterprise-grade ingestion loops over cloud storage ADLS.
• Designed and codified reusable reference architectures for a production-grade agentic AI pipeline leveraging LLMs, LangGraph, OCR, and retrieval pipelines to automate document understanding and decision-making workflows, exposing capabilities through FastAPI services that secured client-facing data platforms via JWT authentication.
• Tariff duty optimization using knowledge graphs and AI agents, generating $3.7 million in revenue impact.
• Optimized LLM inference and agent execution workflows through context engineering, caching strategies, parallel tool execution, and workflow orchestration — significantly reducing latency and improving throughput.

Skills: Generative AI, LLM, LangGraph, AI Agents, MCP, Azure, NLP, OCR, Kubernetes, Document Extraction

• Developed and deployed medical image segmentation models for X-ray and CT imaging using U-Net, DenseNet, and advanced deep learning architectures — achieving >90% accuracy on clinical datasets.
• Owned the end-to-end solution delivery of Bluetail.AI, successfully translating clinical discovery and R&D into a production-grade, enterprise-ready diagnostic SaaS platform across AWS and GCP environments. Leveraging computer vision and deep learning models to detect, segment, and classify 20+ radiological conditions from X-rays — for which 2 patents have been granted.
• Engineered multimodal diagnostic pipelines combining text (radiology reports) + vision models for improved disease detection. Built internal safety and quality validation pipelines ensuring model outputs meet clinical safety guidelines.
• Deployed AI assistants for Marketing, HR, and Research teams by leveraging diffusion models, LLM fine-tuning, and workflow automation to streamline repetitive business processes.

Skills: Deep Learning, LLM Fine-tuning, Computer Vision, AWS, GCP, Multi-Agentic AI, LangChain, Hugging Face

• Fine-tuned and optimized 20+ Large Language Models (LLMs) for diverse enterprise use cases, leveraging techniques such as LoRA, QLoRA, instruction tuning, and domain adaptation.
• Architected scalable RAG and multimodal AI systems capable of ingesting, indexing, and reasoning over text, images, PDFs, and other unstructured enterprise data sources.
• Built ML pipelines for automated model training, evaluation, deployment, and monitoring — improving reproducibility and operational efficiency.
• Architected and delivered key modules for a centralized enterprise Gen AI platform enabling unified access, deployment, benchmarking, and management of multiple open-source foundation models.

Skills: Generative AI, LangChain, LlamaIndex, RAG, NLP, LLMs, Linux, AWS, KFP Pipelines, Python

• Worked on industrial projects for detecting defects on equipment using computer vision. Developed generative AI models and projects related to time series analysis, SQL, and machine learning applications.

Skills: Object Detection, Computer Vision, Machine Learning, Python, SQL, Time Series Analysis

• Created automated marketing AI assistants for use cases like social media post generation, newsletter generation, blog generation, document chatting, and HR AI assistants. Performs multi-agentic RAG to generate articles, blogs, and social content based on company data.

Tech stack: LLM, Diffusion Models, Flux, Stable Diffusion, NLP, Python, GCP, RAG, Prompt Engineering

• Created an agentic WhatsApp assistant connected to a patient database — patients can track medical records and vital readings on mobile via chat. Multi-agentic AI framework. Fine-tuned Llama 3.2 Vision on radiology datasets for disease-specific findings.

Tech stack: LangGraph, Phidata, CrewAI, Unsloth AI, SFT Trainer, MongoDB Atlas, GCP, Python

• Semantic segmentation of deterioration and cracks from concrete datasets using DeepLabv3+ (encoder–decoder architecture) with a ResNet backbone, and video segmentation using YOLO-v8.

Tech stack: Image Processing, Convolutional Neural Networks, Python, MATLAB, TensorFlow, Image & Video Segmentation, PyTorch

• Machine Learning Teaching Assistant, IIT Bhubaneswar — 2021–2023
• Led a team of 20 members organizing various NSS activities — 2017–2021
• Funding team head — successfully raised sponsorship for organizing the college fest — 2019
• School Head Boy (Narayana E-Techno School) — 2014–2015
• Cricket team captain, Narayana E-Techno School, Rajampet — 2013–2015

• Deploying RAG Pipelines to Production at Scale — NVIDIA DL Institute
• Generative AI with LLMs — NVIDIA Inception
• Machine Learning & Deep Learning Masters — iNeuron
• AWS Generative AI Certifications