OPEN TO OPPORTUNITIES

Priyam
Kakati

I am a

Transforming complex data into intelligent action. Expert in building Multi-Agent Systems, RAG Architectures, and production-grade Generative AI solutions.

Top 1%

Global Kaggle Expert

4+

Years Experience

60k+

Learners Mentored


Featured Projects

A mix of Agentic AI, Generative AI, Computer Vision, and Classical ML solutions.

AGENTIC AI (Lead)

MCP Autonomous Recruiter Agents

Sophisticated **Multi-Agent System** using **AutoGen** integrated with Model Context Protocol (MCP). Agents autonomously handle complex workflows: resume parsing, candidate enrichment, and recruiter Q&A. Reduced recruitment MTTR by 25%.

AutoGen MCP GPT-4 LangChain

Enterprise RAG Pipeline

Scalable Retrieval-Augmented Generation system using **Pinecone** and **Weaviate** for vector storage. Implemented advanced chunking and BGE embeddings, boosting document Q&A accuracy by 35%.

Pinecone LlamaIndex RAG

Large-Scale Object Detection

Production deployment of fine-tuned YOLOv5 and Mask R-CNN models for industrial defect detection. Deployed using EC2 instance managing 1000+ parallel inference jobs.

YOLOv5 MLOps AWS

Intelligent Document OCR/LLM

High-accuracy data extraction system combining **OCR** and large LLMs (GPT-4/Claude) for invoice and claims processing. Achieved 80% operational efficiency gain.

OCR GPT-4/Claude AWS Lambda

HFT Time-Series Predictor

Time-series forecasting model using **XGBoost/LightGBM** to predict short-term price movements in a simulated high-frequency trading environment. Optimized for latency.

XGBoost Time Series Pandas/Numpy
AGENTIC AI

Financial Research Agent Swarm

Developed a swarm of specialized agents (Planner, Researcher, Editor, Critic) that collaborate to generate detailed, cited market research reports automatically. Utilized dynamic tool calling.

LangChain Agents Claude 3 Tool-Use Autonomous Workflows

Real-Time Lane Detection

Implemented a fast Deep Convolutional Neural Network (DCNN) for real-time lane detection in self-driving car simulation environments, focusing on inference speed optimization.

PyTorch OpenCV Latency Opt.

Zero-Shot Text Classification

Developed a zero-shot text classification API using pre-trained **Hugging Face Transformers** (BART/NLI models), enabling classification on unseen labels without re-training.

Transformers NLP Zero-Shot

Customer Churn Prediction API

Built a complete ML pipeline for churn prediction using a **Scikit-learn** ensemble model. Deployed as a low-latency microservice via **Flask** with integrated CI/CD on Kubernetes.

Scikit-learn Flask Kubernetes
LLM Fine-Tuning

Custom Instruction Fine-Tuning (QLoRA)

Successfully fine-tuned the **Llama 3 8B** model for a specific industry compliance use case using **QLoRA (PEFT)**. Resulted in a 40% reduction in hallucination rate for compliance queries.

Llama 3 QLoRA/PEFT PyTorch Custom Datasets

Technical Arsenal

Generative AI & LLMs

Multi-Agent Systems (AutoGen) RAG Architectures LangChain/LlamaIndex Fine-Tuning (PEFT/QLoRA) Hugging Face Transformers GPT/Claude API T5/Whisper/CLIP

Core ML & Frameworks

PyTorch / TensorFlow / Keras JAX Scikit-learn XGBoost / LightGBM / CatBoost Time Series Analysis Statistical Modeling

Cloud & MLOps Infrastructure

AWS (SageMaker, EC2, Lambda, S3, Glue) Azure ML / Data Factory GCP (Vertex AI, BigQuery) Kubernetes (CKAD) & Docker Terraform / GitOps MLflow / Airflow / CI/CD Grafana / Prometheus / CloudWatch

Computer Vision & NLP

OpenCV / MediaPipe YOLOv8 / Detectron2 Image Segmentation/OCR BERT / Transformer Models spaCy / NLTK / Gensim Text/Intent Classification

Data Engineering & Stores

SQL (Advanced) Apache Spark / Databricks Apache Kafka MongoDB / PostgreSQL / Redis Pinecone / Weaviate (Vector DBs) Data Preprocessing & ETL

Programming Languages & Tools

Python (Expert) C++ / Rust / Go Java / Scala / R / MATLAB TypeScript / JavaScript Flask / FastAPI / gRPC Git / GitHub / GitLab

Career Timeline

AI Scientist

XO Health Inc. (Remote)

Led AI-powered data extraction and automation systems using GPT-4 and Claude. Built scalable ETL workflows reducing manual hours by hundreds/month.

Data Scientist

Korn Ferry (Remote)

Architected RAG pipelines (Pinecone + GPT-4o). Built MCP-integrated autonomous agents for recruiter workflows. Implemented ML observability stacks.

Data Scientist

Gravitas.ai (Remote)

Developed conversational AI agents (Lex + LangChain) for 10k+ daily users. Established CI/CD pipelines reducing release cycles by 50%.

Data Scientist

NEXTGEN Innovation Labs (Remote)

Trained **YOLOv5** and **Mask R-CNN** models (92% mAP) for Computer Vision solutions. Implemented MLOps practices for scalable deployment and integrated models with **MongoDB** and **BigQuery**.