Welcome to my universe

PrakharBhardwaj

|

Teaching machines to see, learn, and do it better next time. From CMU research labs to production AI systems at BNY Mellon and beyond.

View ProjectsGitHubLinkedIn
Scroll
About Me

Building AI that matters

I'm a Mechanical Engineering MS graduate from Carnegie Mellon University (GPA 4.0/4.0) with a deep focus on Machine Learning and Computer Vision. My journey spans from academic research to building production systems at companies like BNY Mellon and Enterprise Solutions.

I specialize in Large Language Models, RAG architectures, and computer vision pipelines, turning cutting-edge research into scalable, production-ready solutions.

4.0
GPA
CMU
MS Mechanical Engineering
Aug 2022 - May 2024
9.3
GPA
Thapar Institute of Engineering and Technology
BE Mechanical Engineering
Aug 2017 - June 2021
5+
Years in AI/ML
10+
Projects Shipped
Career

Where I've worked

AI Engineer

Enterprise Solutions Inc.

Dec 2024 - Present
  • ▸Built a Chat Assistant web app for Loudoun County Dept. of IT using Cosmos DB, Python backend, React+Vite frontend deployed on Azure Web App.
  • ▸Integrated voice features and user memory personalization to enhance accessibility and conversational depth.
AzureCosmos DBReactPythonLLM

Data Scientist

BNY Mellon

Sep 2024 - Dec 2024
  • ▸Created RAG-based chatbot using on-prem Llama3.1-70b for remediation advice on vulnerable drivers: 88% accuracy, F1 score 0.85.
  • ▸Built RESTful API with FastAPI and integrated it into the frontend, decoupling logic for scalability.
RAGLlama 3.1FastAPILangChainPython

Research Associate

Carnegie Mellon University

May 2024 - Sep 2024
  • ▸Built an end-to-end AI podcast generation pipeline using Llama3 and GPT-4 for script generation (88% accuracy).
  • ▸Integrated voice-cloning text-to-speech model to produce realistic, engaging audio outputs.
Llama 3GPT-4TTSPythonGenAI

Machine Learning Intern

Skylark Labs Inc.

June 2023 - Dec 2023
  • ▸Developed self-learning label management module using YOLOv8, DreamSim, Swin Transformer, and BoT-SORT with 40% accuracy boost via synthetic data generation.
  • ▸Applied super-resolution techniques to sharpen image clarity, achieving 30% improvement in fine-object detection.
YOLOv8Computer VisionPyTorchSynthetic Data

Automation Engineer

Dana Anand India Pvt. Ltd.

July 2021 - June 2022
  • ▸Built deep learning quality control system for bearing cups using Basler ace camera + Jetson Nano, achieving 90% reduction in manual sorting.
  • ▸Increased production efficiency by 25%, scaling output to 100,000 parts/month.
Deep LearningJetson NanoOpenCVMLOps
Work

Things I've built

Robinhood AI Portfolio Copilot

Full-stack AI investment platform with Robinhood/CSV brokerage import, a conversational PydanticAI agent (Gemini 2.5 Flash) over FastMCP tools, portfolio health scoring, macro analysis, real-time stock data, and streaming SSE chat. Full auth, rate limiting, Logfire observability, and Docker Compose deployment.

PydanticAIFastMCPNext.jsFastAPIPostgreSQLGemini

End-to-End AI Generated Podcast

Full pipeline to generate complete podcasts using generative AI, from topic to final audio with voice cloning.

Llama 3GPT-4TTSPython

RAG Chatbot

RAG system using LangChain, Llama 3, Chroma vector DB and OpenAI embeddings, deployed as a Chrome extension.

RAGLangChainLlama 3FlaskReact

Self-Learning AI System

Label management module with YOLOv8, DreamSim & BoT-SORT for dynamic classification and tracking in a self-teaching pipeline.

YOLOv8BoT-SORTComputer Vision

More Projects

3D Printing Defect Detection

PythonDeep Learning

Neural Style Transfer

PyTorchCNN

Image Generation Using GANs

PyTorchGAN

Instant3D: 3D Reconstruction

Computer Vision3D Reconstruction

BERT Fine-Tuning

BERTPyTorch

YouTube Video Summarizer

LLMReact

Foreign Object Debris Detection

YOLOv8ViT
Expertise

My toolkit

Programming

PythonC++SQLRJavaScriptTypeScript

ML Frameworks

PyTorchTensorFlowKerasHugging FaceONNXPEFT/LoRA

LLM & Agents

LangChainLlamaIndexPydanticAIFastMCPOpenAI APIGemini 2.5WandBStreamlit

Computer Vision

OpenCVYOLOv8TensorRTPILSwin Transformer

Web / APIs

FastAPIFlaskNext.jsReactshadcn/uiSSEREST APITanStack Query

Databases

PostgreSQLFaissPineconeChromaCosmos DB

MLOps & Cloud

DockerDocker ComposeMLflowDVCCI/CDAWSAzure AI StudioGCPLogfire

Data & Finance

Finnhub APIyfinancerobin_stocksRechartsZustand
Let's Connect

Got a project in mind?

Whether it's building the next AI system, a collaboration, or just talking shop about LLMs, let's connect.

ai.prakharb@gmail.com