Designed and implemented full-stack architecture for a multi-tenant SaaS platform transforming raw CVE data into actionable intelligence through AI-driven risk scoring.
Built Next.js 16 frontend with React 19, TypeScript, TailwindCSS, and shadcn/ui components with real-time dashboards and dark/light theme support.
Developed FastAPI backend with multi-tenant isolation using tenant-per-database model, implementing RBAC and audit logging services.
Built Celery + Redis worker system to handle asynchronous VRM pipelines at scale, optimizing with Eventlet achieving 200 greenlets per worker for multi-tenant fairness.
Replaced Ollama with VLLM for large language model inference, generating risk narratives and semantic embeddings for CVE mapping to NIST controls.
Enabled CVE similarity search, control mapping, and automated threat analysis using VLLM embeddings.
Designed MongoDB schemas for platform and tenant-specific scan results with tenant isolation, sharding, and replication.
Configured auto-scaling on AWS ECS Fargate, achieving real-time processing for large scans (500 CVEs in ~12-18 mins) with fair task distribution.
Implemented TLS 1.3, PII encryption at rest, RBAC enforcement, and audit logging with NIST 800-53, GDPR, and HIPAA compliance frameworks.
Set up structured logging, performance metrics, and health checks with backup strategy using MongoDB snapshots and Redis persistence.
Engineered PIVORA, an AI-powered document intelligence platform, from concept to deployment, integrating fine-tuned Vision-LLMs (LoRA, RLHF) and developing intuitive React-based dashboards for real-time analytics.
Boosted model accuracy by 24% and data extraction efficiency by 31% via advanced RLHF and transfer learning on 50K+ image-text pairs.
Optimized inference pipelines on AWS S3/Lambda, achieving a 40% reduction in inference cost and a threefold increase in data throughput.
Spearheaded UI/UX implementation utilizing React.js, delivering responsive, reusable components and seamless API integrations for dynamic data visualization.
Designed and deployed responsive dashboards and data visualization interfaces using React.js, React Hooks, and React Router for real-time insights.
Developed reusable UI components and consistent design patterns across the application, improving maintainability and reducing development time by 35%.
Optimized SQL queries and database indexing strategies, reducing execution times and improving data retrieval efficiency.
Ensured software security through OAuth2, JWT authentication, and SAML-based access control.
Collaborated with product designers and backend engineers in an Agile (Scrum) framework to iterate on features and enhance performance.
Implemented robust error logging and monitoring solutions using AWS CloudWatch and Dynatrace, significantly improving system observability and reducing mean time to resolution for critical issues.
Integrated third-party APIs and services to expand platform capabilities and data sources, enabling comprehensive document intelligence and analytics.
Pioneered GPT-4 integration into the RadicalX platform, significantly boosting user engagement by 19% through enhanced platform interactivity, showcasing expertise in Large Language Models (LLMs) and seamless platform integration.
Engineered and deployed NLP-driven recommendation engines, achieving a 25% increase in personalization accuracy across diverse learning modules by leveraging advanced Natural Language Processing (NLP) and recommender system techniques.
Executed in-depth research and sophisticated fine-tuning of domain-adapted LLMs using LangChain and OpenAI APIs, with a critical focus on ethical deployment, performance optimization, and responsible AI principles.
Contributed substantially to Responsible AI initiatives under Dr. Feng Yunhe, encompassing rigorous experiment design, comprehensive model evaluations, and the implementation of robust fairness and safety protocols within AI systems, demonstrating a strong commitment to ethical AI development.
Utilized PyTorch and Hugging Face Transformers for model development and fine-tuning, ensuring optimal performance and ethical considerations.
Collaborated on interdisciplinary teams to address complex challenges in AI ethics and societal impact, translating research into practical applications.
Developed and evaluated AI systems for security, privacy, accountability, and transparency, ensuring adherence to responsible AI guidelines.
Applied deep learning and transfer learning techniques for model generalization and efficiency in various Responsible AI applications.
Technologies:GPT-4LangChainOpenAI APIPyTorchHugging FaceNLPResponsible AI
March 2022 – December 2022
Associate Product Engineer (R&D-product)
TEMENOS – A Banking Software Company
Played a key role in the development of Temenos Infinity Spotlight, a browser-based console for data-driven configuration and seamless integration with core banking systems.
Led development of critical UI/UX features, focusing on user-centered design and cross-browser compatibility.
Collaborated in a high-performing 10-member agile R&D team, participating in sprint planning, rigorous code reviews, and efficient deployment processes.
Resolved complex front-end issues and deployment bugs, improving system stability and reducing UI-related defects by 20%.
Diagnosed and rectified critical front-end issues during deployment phases.
Strengthened product usability through iterative feedback cycles and data-informed enhancements.
Designed and developed scalable solutions for complex business problems, improving system reliability.
Conducted code reviews and implemented best practices for version control using Git and Bitbucket.
Collaborated with stakeholders to gather requirements, analyze business needs, and design optimal software solutions.
Recommended and implemented system enhancements, reducing technical debt and improving maintainability.