Catherene Joshi

AI/ML Engineer with around 3.5 years of experience in developing and deploying AI/ML solutions using AWS, Azure, and GCP. Skilled in NLP, LLMs, and cloud-based architectures. Proven success in delivering scalable AI-driven applications.

AI Engineer at Nallas Corporation (2024-11 – Present)

Creating Global Search feature that searches across multiple product repositories. Developed an AI-powered solution to query multi-indexes each containing documents of different kinds of data. Set up data pipelines and have implemented guardrails, entities extraction, metadata filtering, summary and evaluation functionality.

• Multi-Repo Global Search: Problem Statement: Creating Global Search feature that searches across multiple product repositories
• Solution: Developed an AI-powered solution to query multi-indexes each containing documents of different kinds of data. Set up data pipelines and have implemented guardrails, entities extraction, metadata filtering, summary and evaluation functionality.
• Tech Stack: Cloud Services: AWS Opensearch, Lambda, DynamoDB; Libraries/Frameworks: FastAPI, Langgraph, Spacy, Observability: Opentelemetry, Langfuse, Version Control: Git, LLM: AWS Bedrock: Claude, Nova Lite, Mistral, Llama, Guardrails, Cross-Encoder models, RAG architecture

ML Engineer at Blazeclan Technologies (2023-07 – 2024-10)

Built an AI-driven chatbot with AWS Bedrock using RAG architecture to handle queries addressing cultural differences in patient treatment, restrict responses involving PII, and manage large datasets efficiently while retaining conversational history. Optimized queries via LCELs and FAISS for vector embeddings stored in S3. PII filtering was managed by the LLM, and conversational history retention ensured seamless, context-aware interactions.

Chain of Programming, scratchpad techniques were implemented.

• Document-Based LLM-Powered Chatbot using AWS Bedrock: Problem Statement: The company needed a scalable chatbot to handle queries addressing cultural differences in patient treatment, restrict responses involving PII, and manage large datasets efficiently while retaining conversational history.
• Solution: Built an AI-driven chatbot with AWS Bedrock using RAG architecture, optimizing queries via LCELs and FAISS for vector embeddings stored in S3. PII filtering was managed by the LLM, and conversational history retention ensured seamless, context-aware interactions. Chain of Programming, scratchpad techniques were implemented.
• Tech Stack: Cloud Services: AWS Bedrock, Lambda, S3, DynamoDB; Libraries/Frameworks: LangChain; Architecture: RAG; Vectorstore: FAISS

Cybersecurity Researcher at Cyberkalki Security Pvt Ltd (2022-03 – 2022-06)

Conducted security evaluation of client servers to identify vulnerabilities and ensure compliance with industry standards. Performed VAPT using Nessus, nmap, Burp Suite, and zaproxy to uncover vulnerabilities and misconfigurations. Followed CIS Benchmark for configuration audits to enhance system security.

• Problem Statement: The client needed a security evaluation of their servers to identify vulnerabilities and ensure compliance with industry standards.
• Solution: Conducted VAPT using Nessus, nmap, Burp Suite, and zaproxy to uncover vulnerabilities and misconfigurations. Followed CIS Benchmark for configuration audits to enhance system security.
• Tech Stack: Tools: Nessus, nmap, Burp Suite, zaproxy; Standards: CIS Benchmark

Penetration Testing Intern at Virtually Testing Foundation (2021-10 – 2021-12)

Identified vulnerabilities in web applications to enhance security. Performed VAPT using Burp Suite and Nessus, producing reports on vulnerabilities and remediation strategies.

• Problem Statement: Needed to identify vulnerabilities in web applications to enhance security.
• Solution: Performed VAPT using Burp Suite and Nessus, producing reports on vulnerabilities and remediation strategies.
• Tech Stack: Tools: Burp Suite, Nessus, Metasploit, OWASP ZAP, SQLMap

ML Intern at Grroom (2021-07 – 2021-08)

Developed an application that accurately recognizes fashion items from images. Conducted data collection, preprocessing, feature engineering, annotation, and model training for various ML models, specifically utilizing the YOLO CV model for object detection in images.

• Object Detection for Fashion Items: Problem Statement: The company aimed to develop an application that accurately recognizes fashion items from images.
• Solution: Conducted data collection, preprocessing, feature engineering, annotation, and model training for various ML models, specifically utilizing the YOLO CV model for object detection in images.
• Tech Stack: Google Colab; Libraries/Frameworks: YOLOv8, PyTorch; Annotation Tool: LabelImg

Bachelor of Computer Engineering in Computer Engineering – AP Shah Institute of Technology (2019 – 2023)

HSC – TP Bhatia College of Science (2017)

SSC – Mary Immaculate Girls' High School (2015)