Home

Sankat - AI Architect and Python Developer

[email protected]

Location: Redwood City, California, USA

Relocation: Yes

Visa: USC

Resume file: Sankar - Gen AI Architect_1754420232477.docx Please check the file(s) for viruses. Files are checked manually and then made available for download.

SUMMARY: Seasoned software engineer with 20 years of experience driving innovative solutions across diverse industries, including AI, cloud computing, API Gateway, and payment processing. Proven Agile collaborator with cross-functional teams, bridging technical and business objectives. Currently, building MCP, Model Context Protocol servers to support resources and a set of tools for an enterprise business; I also built MCP clients to be associated with LLM using LangGraph agentic workflow. Developed skill sets to build RAG applications using managed services such as Amazon Bedrock, Knowledgebase, Vector DB; I am also confident to build these services in Python using LangChain, Pytorch frameworks. Investigated replacement of traditional Machine Learning with generative AI models, which could generate new synthetic data, provide more robust forecasting and planning, learn from new data sources, and adapt to changing business scenarios. As part of training, I am currently building agents for Ground Air Traffic Controller, Tower ATC and Pilots for their voice communication using voice to text and text to voice Google API on LangChain platform with Swarm Architecture of LangGraph framework. The workflow is orchestrated and automated to support human Air Traffic Controllers. In Swarm Architecture the state is maintained and could be shared with multiple agents. For instance, the customer information could be shared with Flight Booking Agent as well as with Hotel Booking Agent. At PepsiCo, I was responsible for building the pipeline for data flow from silver layer to gold layer; learning workflow for scenario generation, demand forecasting, and financial evaluation. At the end of the optimization workflow, the model created the planning cycle promotional calendar in excel format with PPG, Pricing and Promotion information for a twelve-week cycle. We deployed the read and download Planning Cycle at Azure API Gateway and created Planning Cycle and model training, data flow was protected within VPN. Comfortable and with confidence, I could integrate enterprise data and applications with evolving Gen AI and Learning in a reliable and scalable way using relevant frameworks, such as Fine Tuning, RAG, Agentic AI or MCP, based on use cases. SKILLS: Financial and Banking: JP Morgan, Key Bank merchant account onboarding, Lowe's implementation of PCI, payment card industry compliance Natural Language Processing: at High wire, used Stanford NLP for content indexing. PepsiCo Gen AI and RAG, built word2vec for embedding. Machine Learning: Linear Regression, Logistic Regression, k-Nearest Neighbor Generative AI: Natural Language Processing, Learning, Named Entity Recognition, Intent Awareness and Sensing, Large Language Model, Reinforced Learning, Fine Tuning, Demand Forecasting, Financial Evaluation, Model Context Protocol Tools and Technologies: Stanford CoreNLP, LangChain, LangGraph, Hugging Face, Meta AI, Oracle Database 23 ai, Generative Pre-Trained Transformer, Chunking and building Knowledgebase, RAG (Retrieval Augmented Generation), Graph RAG, MCP Techniques: Classification, Linear Regression, Scenario Generation, RLHF (Reinforced Learning with Human Feedback), SFT on Vertex AI platform, MCP for integration of Resources and Tools with Microservices Programming Languages: Java, Python Frameworks: MCP, TensorFlow, PyTorch, Hugging Face, LangChain, LangGraph Spring Boot, Fast API, LangFuse Platforms: Kubernetes, AWS Bedrock, Sagemaker, Microsoft Azure Cloud, Google Cloud Platform (GCP), API Gateway Cybersecurity: Built secured applications using Fortify and Sync tools Other: Microsoft SQL Server, Red Hat Drools, Google gRPC, Apache Solr Data Engineering: Market Data Processing: At ValueStox-AI, every night, after the market close, we retrieve the financial data, classify them by sector and industry; we also validate the volatility of each stock and classify them as high volatile, medium and low volatile; that would help investors, especially the Listed Options to decide to invest on PUT or CALL options Competitor s Sales Data: At PepsiCo, for Planning Cycle, we process previous year s sales, marketing and financial data; those are well structured. We also process competitors promotion, pricing and sales data; those are not structured and have missing fields; we enhance the data with available information Enterprise Information: At JP Morgan, every night, as well as on demand, we generate reports for senior management; that involves huge volume of data processing; we adopted ETL and ELT, based on various use cases Digital Publishing Platform: At Highwire, we built digital publishing platform to support 100 publishers and 4000 journals; we persist the metadata of each journal in NoSQL datastores Cassandra and MongoDB; we built multiple channels in Kafka to process high value or high-speed publishing. Domain Driven Design: In the last two decades, many organizations have built microservices based on DDD and CQRS patterns; instead of low-level data, we could effectively handle Products, Customers, Pricing Architecture and Promotional Mechanics. For instance, to promote products for Superbowl event at PepsiCo, we have clear bounded context for data processing. Keywords: artificial intelligence access management database