Position 1: AI System Specialist

Company: UNIS

Period: Mar 2025 - Present

Achievements:

• Architected scalable Gen AI solutions, leveraging LLMs, RAG, and AI agents for enterprise automation, streamlining 10K+ daily operations
• Led a cross-functional initiative with 5+ teams to align AI R&D with key business objectives, achieving a 35% boost in efficiency
• Increased customer-centric retrieval accuracy by integrating RAPTOR, Self-RAG, Agentic hierarchical RAG, and GraphDB-based embeddings
• Built multi-agent systems with semantic and episodic memory allowing collaboration for tasks with 90% success rate

Position 2: AI Graduate Researcher & Teaching Assistant (GenAI)

Company: CSULB Research Foundation

Period: Jun 2023 - Feb 2025

Achievements:

• Optimized LLMs (LLAMA, T5, BERT, Mistral) through transfer learning, PEFT (LoRA, QLoRA), and prompt tuning, boosting downstream task performance by 20% with 70% reduced compute overhead
• Refined LLMs using post-training techniques (quantization, CoT, model distillation) reducing inference time by 30% and improving efficiency
• Developed multi-stage document retrieval integrating BM25 + Cross-Encoder re-ranking to enhance chunk-level ranking & relevance
• Increased customer retention for E-Commerce by 25% through user segmentation with K-Means and NLP models (GPT-4, distilled RoBERTa) achieving 90% pain point identification accuracy
• Partnered with faculty and industry experts, mentored junior researchers to publish research advancing deep learning and generative AI

Position 3: Software Engineer (AI/ML, Full-Stack)

Company: Cognizant Technology Services

Period: Jan 2021 - Dec 2022

Achievements:

• Co-developed AI-powered pilot scheduling optimization system for Alaska Airlines using TensorFlow, PyTorch, and Scikit-Learn, driving 17% increase in operational efficiency
• Developed scalable software applications with Angular, React, and .NET, designing RESTful APIs following industry-standard design patterns
• Optimized SQL queries via advanced indexing strategies and joins, improving system latency by 80% for 300+ operations staff
• Implemented MLOps pipelines, containerization using Docker, and MLflow, ensuring seamless model deployment and monitoring
• Engineered large-scale ETL pipelines with SQL, Pandas, Spark, and AWS Lambda, improving scheduling forecast accuracy by 28%

Programming & Databases

• Python
• C/C++
• Java
• Node.js
• Next.js
• React.js
• SQL
• PostgreSQL
• Neo4j
• Vector DBs (Pinecone, FAISS)

AI & ML

• Generative AI
• LLMs
• RAG
• Agentic AI
• RLVR
• NLP (RNN, LSTM)
• Computer Vision (ViT, CNN, OpenCV)
• Transformers
• Reinforcement Learning
• GANs
• VAEs
• Data Science
• Statistics
• Data Structures & Algorithms
• Machine Learning
• Data Visualization
• AI-powered developer tools
• Self-RAG
• Agentic Hierarchical RAG
• RAPTOR
• GraphDB Embeddings
• Multi-Agent Systems
• Semantic Memory
• Episodic Memory

Frameworks & Libraries

• PyTorch
• TensorFlow
• LangGraph
• Scikit-learn
• Pandas
• NumPy
• Apache Spark
• CrewAI
• Keras
• Hugging Face
• React
• Angular
• .NET
• FastAPI

Cloud & MLOps

• AWS (Sagemaker, Bedrock, API Gateway)
• Vercel Deployments
• Terraform (IaC)
• Docker
• Git
• CI/CD
• MLflow
• MLOps
• OpenAI SDK
• MCP
• Apache Kafka

Development Tools & AI Editors

• Cursor
• Claude Code
• GitHub Copilot
• n8n Low-Code
• AI-Powered Editors
• Code Generation Tools

Other

• Open-source (AI Agents Marketplace, MCP/A2A samples, Research)
• API Development (.Net, FastAPI, RESTful, A/B testing)
• OOD
• SDLC best practices
• Design patterns

Project 1: Unified MCP Tool Graph

Description: Revolutionary AI agent framework that fixes tool confusion by providing structured, intelligent tool selection for LLMs. Instead of dumping 1000+ tools into prompts, it equips agents with clarity and relevance.

Technologies: TypeScript, Python, Neo4j, MCP Protocol, LangGraph, A2A Protocol, GraphQL

GitHub: https://github.com/pratikjadhav2726/Unified-MCP-Tool-Graph

Project 2: MedGPT: AI-Driven Medical Vision Q&A System

Description: Built an end-to-end multimodal LLM by fine-tuning and prompting LLaVA to extract insights from medical images, EHR and clinical texts, ensuring HIPAA & GDPR compliance and delivering GPT-like diagnostic results with 87% accuracy.

Technologies: LLM, Machine Learning, LLaVA, PEFT, LoRA, Medical AI

GitHub: https://github.com/pratikjadhav2726/MedGPT

Project 3: LinkedInApply: Smart Job Application Bot

Description: Automated 500+ end-to-end job applications across LinkedIn, reducing manual effort by 70% through AI-driven RAG based form answering with 90% accurate APPLY/SKIP tokens.

Technologies: AI, RAG, Automation, LinkedIn API, Job Matching, Python

GitHub: https://github.com/pratikjadhav2726/LinkedInApplyAutomation

Project 4: Diffusion-Based Text-to-Speech (TTS) with D3PM for Voice Cloning

Description: Built a SOTA Speech Synthesis system by replacing autoregressive models with diffusion models & D3PM, reducing inference time by 34% and enabling zero-shot voice cloning from a 3-second sample.

Technologies: Generative AI, Diffusion, TTS, D3PM, Voice Cloning, Python, PyTorch

GitHub: https://github.com/csulb-datascience/TTS-with-Diffusion-model

Project 5: AG-UI Chatbot: Generative UI with MCP Server Integration

Description: A Next.js chatbot that uses Large Language Models (LLMs) and an MCP server to generate dynamic, structured UI templates from user requests, instantly creating dashboards, tables, charts, cards, and more.

Technologies: Next.js, TypeScript, LLM, MCP Protocol, Amazon Bedrock, Claude 4 Sonnet, React, AWS

GitHub: https://github.com/pratikjadhav2726/AG-UI-chatbot