Feb 2025 – Sep 2025 · 7 mos · Gurugram, Haryana, India · On-site

• Diagnosed inefficiencies in production data pipelines, pinpointing root causes of data inaccuracies
• Architected and implemented optimized batching service, reducing operational costs by 50%
• Orchestrated deployment of FLUX with ComfyUI on AWS p5.48xlarge infrastructure
• Engineered high-performance image generation pipeline achieving sub-2-second latency and 30 concurrent requests throughput

Dec 2024 – Feb 2025 · 2 mos · Gurugram, Haryana, India

• Architected end-to-end financial infrastructure integrating PhonePe APIs to automate
• subscription management and accounting workflows
• Engineered WhatsApp-integrated conversational AI system, enabling seamless customer
• onboarding and support while significantly reducing operational overhead
• Designed comprehensive quality assurance framework for AI-generated responses, integrating
• systematic feedback loops into administrative workflows
• Established technical recruitment pipeline, conducting technical interviews and capability
• assessments for core AI engineering talent

Apr 2023 – Dec 2024 · 1 yr 8 mos · Noida, Uttar Pradesh, India

• Manage a team of 5 engineers and product managers
• Lead the Generative AI department at Valiance
• Built lucrative, long-term relationships with clients and customers
• Facilitated communication between engineering team and clients to maintain on time delivery of products and services
• Responsible for the upholding the product and service standards beyond client expectations
• Built and Deployed state-of-the-art RAGs with LLMs to solve large scale data based Q&A with focus on accuracy.
• Utilized Google, AWS and Azure Cloud Service for rapid deployment and scaling of solutions.
• Built, Managed and Modified Knowledge Graphs, Vector Stores and other data stores to augment LLMs.
• Established data ingestion pipelines to extract and preprocess data for highly accurate and reliable data retrieval and responses.
• Developed solutions to readily plug and play with any available cloud service or LLM
• Designed a formal evaluation pipeline for comparing the performance of various Q&A pipelines.
• Built a tendor verification automation pipeline to reduce hours of high stakes verification tasks.
• Built a reliable logging and traceback system to allow rapid debugging of errors.

Jun 2022 – Aug 2022 · 2 mos · Gurugram, Haryana, India · Remote

• Designed and deployed a novel recommendation system to find better Google AdWords based on data from the past 4 years.
• Generalised the model over various languages and regions across the globe using translations and
• normalizing the data from different regions.
• Designed an NLP model to create optimal encoding for key ”phrases” from past AdWords data.

Jun 2022 – Aug 2022 · 2 mos · Irvine, California, United States · On-site

• Performed a comprehensive literature survey of Black Box AI Attacks.
• Implemented POBA-GA and Curls and Whey to attack an end-to-end ACC system.
• Implemented model-stealing and adversarial attacks on depth estimation systems.

Jan 2022 – May 2022 · 4 mos · India · On-site

• Innovated novel reinforcement and active learning techniques for model extraction.
• Achieved significant improvement over the current state-of-the-art results in model stealing attacks.
• Implemented KnockOffNets and ActiveThief for benchmarking.
• Implemented Entropy, Monte-Carlo Dropout, k-centre and other active learning based dataset selection techniques for model stealing.
• Executed data augmentation using pseudo-labeled data and other weak and strong augmentation.
• Implemented FixMatch, MixMatch and other Semi-Supervised Learning techniques to increase the efficiency of model extraction

May 2021 – Aug 2021 · 3 mos · Munich, Bavaria, Germany · Remote

• Implemented novel layer-wise abstraction techniques using Linear Programming and Orthogonal Projection
• Implemented counter-example guided gradient descent for refinement of the network.
• Implemented a fully functioning CEGAR loop to use the novel abstraction and refinement techniques for easier verification and proof of the robustness of large neural networks.

Mar 2021 – Apr 2021 · 1 mo · Gurugram, Haryana, India · Remote

• Designed and deployed a web scraper to collect news articles in real-time from the web.
• Analysed the news articles and extracted keywords to find sentiments and keywords relevant to the company.
• Used reinforcement learning to improve weights for keywords, resulting in better results over time.