Oct 2023 – Sep 2025 · 1 yr 11 mos

• Spearheaded the Responsible AI (RAI) initiatives in Alexa+, reducing unsafe model outputs (toxicity, bias, sexual content) through targeted alignment and safety strategies.
• Applied Model finetuning and Retrieval-Augmented Generation (RAG) to fine-tune LLMs for grounded, policy-aligned responses.
• Created an Agentic-AI based prompt optimization framework using feedback and refinement agents to autonomously enhance model performance on RAI benchmarks.
• Scaled RAI solutions into production by collaborating with policy and engineering teams across AGI org.

Mar 2020 – Sep 2023 · 3 yrs 6 mos

• Published research paper on “Learning under label noise for robust spoken language understanding systems” in Interspeech 2022, Incheon, South Korea.
• Led a team of five to complete various Amazon goals, which further resulted in improved accuracy of Alexa across IN locales by 50 bps.
• Successfully delivered on the simultaneous Netflix launch across all Alexa locales, accuracy improvement goals, fairness in AI and robustness on noise research science programs.
• Submitted research publications on fairness in AI, mitigating algorithmic bias through continual learning and noise robustness of IC-NER systems in EMNLP Industry Track 2023.

Sep 2017 – Feb 2020 · 2 yrs 5 mos · Gurgaon, India

• Speech Recognition and NLP:
• Created in-house bilingual Automatic Speech recognition machine learning model for “Hinglish”
• Created GPU based pipeline transcribing more then 100,000 real time calls everyday
• Call center automation for Quality assessment of calls based on neural networks
• Improved revenue per lead by 8% using Machine Learning based Lead propensity.
• Callbot and Whatsapp based Chatbot:
• Automated 70% of post sales calls/chats by using Natural Language processing based conversational bot
• Deep learning based Video Inspection:
• Automated the process of manual inspection for Car Insurance using computer vision
• Pipeline is based on Scratch and dent detection in user submitted videos
• Work published in Techcircle, TOI, ET, AWS blog , Softbank Portfolio presentation

Aug 2015 – Mar 2017 · 1 yr 7 mos · Gurugram, Haryana, India

Nov 2010 – Aug 2015 · 4 yrs 9 mos · Noida Area, India

• Langhar.com was one of first few food tech startups in India. It started as marketplace for homely food, which empowered housewives to earn extra income while working from home.
• Roles and responsibilities-
• ➢ Raised funding of $200,000 from Ark Challenge USA, Amity university and Times internet.
• ➢ Handled Web and mobile app development (YII, MongoDB, AWS EC2 and EBS, Android, iOS)
• ➢ Worked on Delivery cost optimization, User experience design, Digital marketing, affiliate marketing, referrals, third party tie-ups.