Pushkar Kurhekar

Boston, MA, USA

contact@pushkar.io

PDF

Summary

Machine Learning Engineer with 3+ years of experience designing and deploying scalable ML solutions, including fine-tuning large language models, building retrieval-augmented generation (RAG) pipelines, and integrating LLM APIs. Proficient in Python and ML frameworks such as PyTorch and TensorFlow, consistently delivering performance improvements and optimized inference. Expertise includes developing robust evaluation frameworks and streamlining end-to-end ML pipelines for production environments.

Experience

Machine Learning Co-op

CAMP4 Therapeutics Corporation

Jan 2024 - Aug 2024

Engineered and fine-tuned large language models using PyTorch for production-level sequence prediction tasks, integrating advanced NLP techniques to boost accuracy by 25% and reduce inference time by 15%.
Implemented LoRA fine-tuning pipeline for LLM adaptation on 8,000 proprietary documents, achieving 11% F1-score improvement while reducing training time by 70%.
Architected RAG system using Weaviate vector database for semantic search across 5,000+ internal documents, improving retrieval accuracy by 40% and reducing search time from hours to seconds.
Deployed MLflow for systematic experiment tracking and model versioning across three projects, standardizing workflows and reducing performance variance by 20%.
Established ML monitoring with Evidently AI for real-time performance tracking and drift detection, maintaining 90%+ model accuracy through automated alerts and retraining.

Systems Engineer

Tata Consultancy Services

Jul 2019 - May 2022

Developed pre-trained language models on AWS supported by containerized workflows (Docker/Kubernetes) and CI/CD integration, achieving a 35% boost in accuracy and a 40% improvement in efficiency.
Implemented an automated document extraction system using OpenCV and TensorFlow integrated with MongoDB, enhancing accuracy by 64% and speed by 30% while contributing as a patent co-inventor.
Engineered scalable MLOps pipelines using AWS SageMaker, Lambda, ECS, and Kubernetes orchestration to productionize ML models, integrate model serving and performance monitoring, and reduce system downtime by 25%.
Developed multi-class text classification system using spaCy and XGBoost for document categorization across 12 categories, achieving 91% accuracy and processing 7,000+ documents daily via REST API.
Implemented transformer-based semantic similarity system for intelligent document matching using Hugging Face Transformers, reducing manual review time by 70%, deployed on SageMaker endpoints processing 5,000+ comparisons daily.

Education

MS, Computer Science

Northeastern University

Jan 2023 - May 2025

GPA: 4.0/4.0

Bachelor of Engineering, Computer Engineering

University of Mumbai

Jul 2015 - Jul 2019

Projects

Generative AI Email Assistant

AI-Powered Recipe Generator

Food Image Classification

Mobile Price Classification

Real Time Sign Language Estimation System

Machine Learning Co-op

CAMP4 Therapeutics Corporation

Jan 2024 - Aug 2024

Systems Engineer

Tata Consultancy Services

Jul 2019 - May 2022

MS, Computer Science

Northeastern University

Jan 2023 - May 2025

Bachelor of Engineering, Computer Engineering

University of Mumbai

Jul 2015 - Jul 2019

Generative AI Email Assistant

AI-Powered Recipe Generator

Food Image Classification

Mobile Price Classification

Real Time Sign Language Estimation System

Denoising Images with Autoencoders

US Patent 12,056,171 B2 (Co-inventor)

August 2024

Automated Text and Tabular Data Extraction from Scanned Document Images

January 2021

Real Time Sign Language Estimation System

October 2019