Company Overview: Client: Exxon Mobil, Texas, United States
Business Domain: Audit & AI
Hands-on Presales Expertise in AI/ML as I collaborated with major analytics projects to design and implement tailored AI solutions, enhancing client engagements and operational success
Developing a PoC on document question-answering/matching for long audit procedures documents using Langchain, longformer, Milvus, AutoGPT, FAISS & Cohere APIs
A digital coach in audit process which aims to provide semantic search and re-ranking in Audit guidance documents using Cohere, Azure Openai, Azure Cognitive Search
A PoC on Auditor performs risk assessment of an entity's business using various industry internal and external factors to identify overall financial position using Azure Openai, Azure Cognitive Search & Falcon 7B models
Scrapping input from Audit guidance standards like FASB, SEC Regulations, AICPA
Configure API keys & preprocess input documents like read, split/chunking
Storing and indexing vectors in vector DB - Azure CognitiveSearch & FAISS
Vector embeddings using sentence embeddings like Ada, Davinci, SentenceBERT
Document matching, re-ranking documents
Data summarization using LLM models GPT-3.5 turbo, GPT-4 & Falcon 7B
Company Overview: Client: One Main Financial (Bank) in Buffalo, NY
Business Domain: Banking, Auditing, IoT, Healthcare, Insurance & Pharmaceutical
Build Supervised and Unsupervised traditional AI models
Understanding client requirements, mapping problem definitions with AI/ML solutions
Working on RFPs, PoCs and MVPs, creating roadmaps, architectures, strategies to develop AI solutions & active solution review meetings
Day-to-day interactions with end clients, leading client teams & project deliveries
Worked with other partner clients on EHR, claims and health insurance datasets
Designed a block level technical architecture & facilitate in architecture
Developed & deployed a Sharecare QnA & semantic search chatbot using Azure CognitiveSearch, Formrecognizer, Azure OpenAI Studio, Azure Devops components
Developed an MVP on medical document parser for tables and text extraction in fax scan documents from MGB channel DB which is created on Azure Databricks using AWS Comprehend, Azure Formrecognizer & deployed it in Azure Devops
Assesment of denoising & two-stage object-detection models like RCNN variants
Testing feasibility of custom OCR models and fine-tuning BERT model variants
Validating state of the art transformer based NER models and fine-tuning with few-shot & zero-shot learning models
Develop & test various custom NLU pipelines and validation using customized BERT
Creating intents, rules, stories and slots and running validation experiments using Dialogflow, Amelia and RASA NLU