ANTRIKSH PANDEY

• Adoption of 10M+ users across servicenow generative ai platforms with core LLM models,integrations & workflows.
• Built LLMOps platform with Model Finetuning & inferencing for NER with DistilBERT,knowledge distillation, summarization with T5,LLAMA base for chat summarization tasks .
• Developed Core RAG Framework for servicenow with data sources integration,data preprocessing, embeddings as a service & application integration.
• Commercialization with "Assists" as a subscription model , driving first 1Million+ users for core Now assist platform.
• Developed key layers for Trust and safety of LLM across hallucinations, offensiveness, prompt injection & data privacy for servicenow.
• Model Pruning & Quantization: Leading the effort for model experimentation with quantization efforts(on GGUF &LLama.cpp),model parallelism(data,pipeline & tensor), model optimization strategies(KV cache, continues & speculative batching).
• Generative Ai agents development: Driving core platform initiative with skills,models,agents and workflow for core use cases in sequential problem solving.

• Generative AI Platform(GenAAS): Conceptualized Generative Ai platform, suite of managed services for designing & testing prompts, tuning generative AI model with inbuilt LLM gateway, embeddings generator,Vector DBs,Open AI & Google Ai models.

• Deep ETA: Built 0-1 the platform experiences for supply chain ETA platform for receiving time with tuning across 7 different neural network architectures resulting in improvement across latency,accuracy & GA for ETA models.

• Conversational Commerce(CC): Researched and conceptualized CC experiences for customers enabling Voice & Text AI through shopping assistant-based capabilities.

• Established strong relationships with key stakeholders, fostering collaboration across departments for seamless product development.

• Conversational Experience Platform (ConNLU):Conversational AI platform exposed as an open API with pluggable channels, intent recognition, and intent handling capabilities, allows customers to interact with Jockey using natural English language to order online grocery, in addition to several other use cases like customer care, returns, etc. In general, the multi-channel-multi-tenant design paradigm of the platform allows developers to build and deploy any conversational application and serve customers on any arbitrary client like Google Assistant, Siri, Unix shell, chrome plugin, etc.

• Quote To Cash Revenue Platform (SAAS): Defined the key Quote to Cash journey capabilities-Building sales quotes with product modules like price & quote, order renewals, product configurator, guided selling.
• Proposal & Contract Management: Defined as part of Quote to cash process product modules like proposal generator, manage contracts.
• Revenue & Subscription Management: Integrating the invoicing, collections, taxes, and reporting for enterprise businesses through modules like invoicing, subscription billing, payments, taxes & VAT, revenue recognition.
• Revenue intelligence (Machine Learning)- Built the “Augmented ML engine” supporting Revenue forecaster along with insights module

• Collect, Index & Query: Conceptualized the core omnichannel ecommerce engine to power search results through Apache SOLR.
• Search Ingestion All upstream systems which feed data to search integrate with the search ingestion services for indexing
• Search Relevance: Query Intent: Understand more about the query including attribute tagging, spell check and other normalization.

• Customer Intent: Understanding more about the customer for the context. Owned by personalization team and provided P13N service for each markets.

• ReRanking: These ranking services, reranks the top N products to the best order, that customer is likely to click and purchase. Includes personalized ranking for grocery.

• Item engagement signals: These signals eg, CTR, trending, order rate, store order rate are calculated based on customer engagement on the items overall and to the context that customer is interacting.
.

• Product Development: Designed & implemented SMART Automation Systems through data collection, data source optimization & cloud analytics on Google Cloud Platform (GCP)
• Traction Automation: Designed solutions in traction automation system through marketplace estimation, product use, specifications & manufacturability; Conducted tolerance testing through component optimization.