Portfolio

Yash Khairnar

AI/ML engineer and CS graduate student at SJSU focused on multi-modal AI systems. Welcome to my space.

Yash Khairnar
GitHubLinkedInCVEmail

Research / Coursework

Interests

  • -World Models
  • -Core ML/DL, Computer Vision, LLMs, RL and Multimodal systems
  • -Agentic AI Systems
  • -Retrieval-Augmented Generation
  • -Game Theory

Selected Coursework

  • -Machine Learning, Deep Learning, Computer Vision, Natural Language Processing, Reinforcement Learning
  • -Distributed Systems
  • -Data Structure & Algorithms
  • -DBMS, OS, Computer Networks
  • -Cloud Computing, Software Engineering

Education

San Jose, CA

M.S. Computer Science

San Jose State University

Graduate study focused on AI/ML systems, distributed systems, and applied software engineering.

Pune, India

B.E. Computer Engineering

Savitribai Phule Pune University

Undergraduate foundation in computer engineering, algorithms, databases, systems, and applied machine learning.

Projects

Documentation Assistant

case studycode

RAG application that indexes technical documentation URLs and answers questions with retrieved context.

Python, FastAPI, LangChain, RAG

Nook

sitecode

Cafe discovery platform with personalized recommendations, location-aware search, and a mobile/web product surface.

Next.js, React Native, AWS, collaborative filtering

ThinkFlow

case studycode

EEG-to-text translation system using VQ-VAE, BART Transformer, and LSTM Seq2Seq with attention.

PyTorch, Hugging Face, Flask, Next.js

Sentinel

case studysitecode

Serverless email marketing platform with analytics, AI-assisted content generation, and multi-region AWS deployment.

AWS Lambda, S3, API Gateway, DynamoDB, SQS, SNS, SES, Terraform

Malware Classification

code

Classified malware into 12 families using DenseNet121 CNNs on 10K+ samples, reaching 95% accuracy.

PyTorch, DenseNet121, SVM, grid search

AetherForge

case studydemocode

Multi-agent collaborative IDE where humans and AI agents can edit the same codebase with CRDT-backed sync.

CrewAI, Gemini orchestration, Automerge, React

Writing

Performance Analysis: LSTM vs RNN

Performance Analysis: LSTM vs RNN

Comparative study using Google Speech Commands dataset. Understanding the trade-offs between architectural choices in recurrent networks.

medium

Experience

Jul 2024 - Jun 2025

Software Engineer, Accurate Industrial Controls Pvt. Ltd.

  • -Engineered an end-to-end predictive maintenance system for industrial generators using anomaly detection and RUL prediction models, achieving 93% accuracy across 500+ hours of telemetry data.
  • -Built an Automatic Number Plate Recognition pipeline using YOLOv11 and PaddleOCR with image preprocessing modules, improving plate detection and text extraction to 97% accuracy.
  • -Designed a ROS-based inspection pipeline integrating YOLO, EasyOCR, and NVIDIA Triton Server, eliminating manual intervention in LPG cylinder inspection with 93% detection accuracy.
  • -Deployed optimized ML pipelines with ONNX and Triton, reducing inference latency by 40% and increasing throughput for real-time systems.

Aug 2023 - Nov 2023

Artificial Intelligence Intern, Accurate Industrial Controls Pvt. Ltd.

  • -Built an anomaly detection system for copper coil inspection using PatchCore and YOLO, integrating object tracking and image compression to improve inference efficiency.
  • -Reduced video inference latency to 50 ms using GPU acceleration, multi-threading, frame skips, and pipeline-level optimizations.
  • -Evaluated and implemented classical AI search algorithms including A* and potential fields for autonomous navigation.

Nov 2022 - Feb 2023

Deep Learning Intern, ResoluteAI Software

  • -Developed a facial recognition attendance system using MTCNN for detection and a custom ANN classifier for embeddings, achieving 90% accuracy with less than 2% FPR.
  • -Engineered a PDF resume parser using PyPDF2 and regex to extract structured data into JSON, automating 70% of manual data entry.

Project Archive

Agies

code

AI-powered automated bug-fixing agent that analyzes Sentry errors, identifies problematic code, proposes fixes, tests them in a sandbox, and creates draft pull requests.

Agentic AI, LangGraph, Streamlit, Sentry, CodeRabbit, Daytona Sandbox, GitHub API

CuriosityAI

democode

Agentic research system that identifies innovation gaps using embedding inversion and retrieval-augmented generation, producing invention abstracts, sketches, and proposals.

CalHacks 12.0, agentic AI, voice agent, embedding inversion, RAG

ResuMatch

democode

RL-based resume optimizer using a two-agent architecture with an editor and judge, plus reward shaping for tailoring resumes to job postings.

Next.js, Tavily, reinforcement learning, LangGraph, Flask

Slique

sitecode

Hiring platform connecting fashion models with brand recruiters through job posting, search, applications, real-time chat, and contract-based hiring.

Next.js, Tailwind CSS, FastAPI, PostgreSQL, AWS RDS

Image Colorizer

code

Deep learning app for colorizing black-and-white photographs using a GAN-based image translation workflow.

Flask, HTML/CSS, NumPy, OpenCV, GANs, model fine-tuning

4Bit

code

Diagnostic tool for lung cancer analysis using multi-omics data, CT scan images, biomarkers, and genetic mutation signals.

Python, Keras, TensorFlow, multi-omics, biomarkers

Roomie Radar

code

Roommate matching app that uses preference-based compatibility scoring to recommend better roommate fits.

Next.js, machine learning

Contact

Email is the best way to reach me.

GitHubLinkedInWritingCVEmail