Purva Chiniya

AI Researcher

Seattle, Washington, United States6 yrs 3 mos experience

Key Highlights

Expert in hyperparameter optimization and NLP.
Developed scalable AI solutions at Amazon.
Published research in top AI conferences.

Stackforce AI infers this person is a skilled AI Researcher specializing in NLP and machine learning applications.

Contact

Skills

Core Skills

Hyperparameter OptimizationNatural Language Processing (nlp)Recurrent Neural Networks (rnn)

Other Skills

Search Engine Optimization (SEO)Data GenerationApplied Machine LearningData EngineeringMachine LearningDeep LearningPerlJavaComputer VisionImage ProcessingMatlabC++oratorlawn tennisNeural Networks

About

At Amazon, contributed to Responsible AI for Alexa+, focusing on ethical and inclusive AI solutions. Partnered with cross-functional teams to refine language models and enhance user experiences. Previously, as an AI Research Intern at Sentient Foundation, worked on hyperparameter optimization and fine-tuning pipelines for deep search and computer use agents. Graduated with a Master's in Computer Science at the University of Maryland with a solid foundation in Electrical Engineering from IIT Roorkee. Motivated by a passion for leveraging NLP and machine learning to drive impactful AI applications, aiming to develop scalable and responsible technologies for diverse global communities.

Experience

6 yrs 3 mos

Total Experience

1 yr 8 mos

Average Tenure

1 yr 2 mos

Current Experience

Amazon

Applied Scientist

Jul 2025 – Present · 11 mos · Bellevue, Washington, United States · On-site

Working on production Agents for Amazon Ads

Sentient foundation

AI Research Intern (Co-op)

Oct 2024 – May 2025 · 7 mos · Remote

Engineered a production-grade hyperparameter-optimization (HPO) and finetuning platform that automatically tuned datamix and training parameters to customer-defined objectives (alignment, factuality, robustness) using adaptive sampling, LR scheduling, and performance-aware data-mix reweighting.
Augmented the HPO system with data-generation pipelines to synthesize finetuning and DPO datasets, enabling continuous dataset expansion and delivering 30% fewer tuning iterations and 40% faster finetuning cycles across 8b and 70b models.
Engineered the reasoning-agent orchestration layer for Open Deep Search (ODS), implementing LiteLLM/SmolAgents integration, a fully reproducible evaluation harness, and config-driven indexing and caching for scalable, self-hostable reasoning search.

Hyperparameter OptimizationSearch Engine Optimization (SEO)

Amazon

Applied Scientist intern

May 2024 – Aug 2024 · 3 mos · Bellevue, Washington, United States · On-site

Developed a lightweight gradient-based controlled decoding mechanism achieving over 50% reduction in false positives, significantly improving system reliability.
Decreased attack success rates by more than 20% compared to baseline models on GradSafe, SFT, and DPO.
Paper accepted at AMLC 2024 and LREC 2026.

Natural Language Processing (NLP)Applied Machine LearningData Engineering

University of maryland

Research Assistant

Aug 2023 – May 2024 · 9 mos · United States · On-site

Working as a research assistant under the supervision of Professor Cody Buntain at Information Studies department

Deep LearningNatural Language Processing (NLP)

Cisco

Software Engineer

Aug 2021 – Aug 2023 · 2 yrs · Bengaluru, Karnataka, India · On-site

Worked in the XR BGP team.

Indian institute of technology, roorkee

2 roles

Machine Learning Lab Intern

May 2019 – Apr 2020 · 11 mos

Secretary, Lawn Tennis Team

Apr 2019 – Jan 2021 · 1 yr 9 mos

University of tennessee, knoxville

Research Intern

May 2019 – Aug 2019 · 3 mos · Knoxville Metropolitan Area

Data Science Internship

Sdslabs

Data Science Group IIT ROORKEE

Feb 2019 – Jun 2021 · 2 yrs 4 mos · IIT ROORKEE

SDSLabs is a student group that constantly tries to innovate and foster technical activities in campus.
The group conduct lectures and workshops on data science topics and also conduct monthly competitions on our various self-developed applications.

Recurrent Neural Networks (RNN)

Indian institute of technology, bombay

Research Intern

May 2018 – Jul 2018 · 2 mos · Mumbai Metropolitan Region

In the team of two, I worked on "Application of deep learning to detect age from image"under Professor Biplab Banerjee of IIT Bombay.
Aim was to implement neural networks on the given image dataset.
Using MATLAB we extracted necessary information and then used tensorflow framework of python to implement the age detection model.