Featured Projects

ASTRA Logo

ASTRA - Security & Defense Platform

Lead Developer | 2024

Developed ASTRA, a comprehensive security and defense platform designed to protect organizations from emerging threats. As the lead developer, I architected and built the full-stack solution with advanced security features, real-time threat detection, and intelligent defense mechanisms.

Key Features

Real-Time Threat Detection

Advanced monitoring and detection systems that identify and respond to security threats in real-time using AI-powered analysis.

Intelligent Defense

Automated defense mechanisms that adapt to evolving threats, providing proactive protection for critical infrastructure and assets.

Security Analytics

Comprehensive security analytics dashboard providing insights into threat patterns, attack vectors, and security posture.

Scalable Architecture

Built with scalability in mind, supporting enterprise-level deployments with high availability and performance.

Technical Implementation

  • Frontend: Built responsive React.js dashboard with real-time data visualization, interactive security maps, and comprehensive reporting interfaces
  • Backend: Developed robust REST APIs using Python/Django with microservices architecture for high performance and scalability
  • AI/ML Components: Implemented machine learning models for:
    • Threat pattern recognition and anomaly detection
    • Predictive security analytics
    • Automated threat response systems
    • Behavioral analysis and risk scoring
  • Security: Implemented end-to-end encryption, secure authentication, and compliance with industry security standards
  • Integration: Seamless integration with security tools, SIEM systems, and threat intelligence feeds
  • Deployment: Containerized deployment with Docker and Kubernetes for scalable cloud infrastructure
React.js Python Django Machine Learning Security REST APIs PostgreSQL Redis Docker Kubernetes AWS Threat Intelligence
Astragar VRM

Astragar VRM - Multi-Tenant Vulnerability Risk Management Platform

Lead Developer (Full-Stack, Backend, AI/ML) | 2024

A SaaS platform for multi-tenant Vulnerability Risk Management that transforms raw CVE data into actionable intelligence through AI-driven risk scoring, compliance checks, and threat intelligence. Designed and implemented the full-stack architecture, from frontend to backend, distributed workers, and AI/ML integrations using VLLM.

Key Contributions

1. Full-Stack Development

  • Built Next.js 16 frontend with React 19, TypeScript, TailwindCSS
  • Implemented shadcn/ui and Radix UI components
  • Responsive, mobile-first dashboards with real-time updates
  • Dark/light theme support
  • Integrated Stytch B2B authentication for passwordless login

2. Backend & API Architecture

  • Developed FastAPI backend with multi-tenant isolation
  • Tenant-per-database model for security
  • RBAC and audit logging services
  • Modular APIs: organization, member, scan, NIST controls, MITRE ATT&CK mapping, security news feed

3. Distributed Task Processing

  • Built Celery + Redis worker system for async VRM pipelines
  • Optimized with Eventlet: 200 greenlets per worker
  • Multi-tenant fairness in task distribution
  • Stage-based pipeline: extract, similarity search, exploit enrichment, NIST mapping, compliance scoring, risk metrics, MITRE scenario mapping, report generation

4. AI/ML Integration with VLLM

  • Replaced Ollama with VLLM for LLM inference
  • Generated risk narratives and semantic embeddings
  • CVE similarity search and control mapping
  • Automated threat analysis using VLLM embeddings

5. Data Architecture & Multi-Tenancy

  • MongoDB schemas for platform and tenant-specific data
  • Tenant isolation, sharding, and replication
  • Immutable audit logs and idempotent task execution

6. Performance & Scalability

  • Auto-scaling on AWS ECS Fargate
  • Real-time processing: 500 CVEs in ~12-18 mins
  • Fair task distribution across tenants
  • Monitoring: API latency, worker task depth, resource usage

7. Security & Compliance

  • TLS 1.3, PII encryption at rest, RBAC enforcement
  • NIST 800-53, GDPR, and HIPAA compliance frameworks
  • Secured against XSS, injection attacks
  • Strict CORS and rate limits

8. Observability & Disaster Recovery

  • Structured logging, performance metrics, health checks
  • Backup strategy: MongoDB snapshots, Redis persistence
  • Auto-scaling, retries, and failover mechanisms

Impact

  • ✅ Enabled actionable AI-driven vulnerability analysis for multiple tenants simultaneously
  • ✅ Delivered real-time dashboards, automated risk scoring, and compliance insights
  • ✅ Built a scalable, secure, and maintainable platform ready for enterprise adoption

Technical Stack

Next.js 16 React 19 TypeScript TailwindCSS shadcn/ui FastAPI Celery Redis VLLM MongoDB AWS ECS Fargate Stytch Eventlet NIST 800-53 MITRE ATT&CK
PIVORA Logo

PIVORA - AI Document Intelligence Platform

SUNUS AI | March 2025 - Present

Engineered PIVORA, an AI-powered document intelligence platform, from concept to deployment. The platform integrates fine-tuned Vision-LLMs using LoRA and RLHF techniques, enabling intelligent document processing and data extraction at scale.

Key Achievements

  • Boosted model accuracy by 24% through advanced RLHF and transfer learning
  • Increased data extraction efficiency by 31% on 50K+ image-text pairs
  • Achieved 40% reduction in inference cost on AWS S3/Lambda
  • Tripled data throughput through pipeline optimization
  • Reduced development time by 35% through reusable UI components

Technical Implementation

  • Fine-tuned Vision-LLMs using LoRA (Low-Rank Adaptation) for efficient model adaptation
  • Implemented RLHF (Reinforcement Learning from Human Feedback) for improved model performance
  • Built React-based dashboards with real-time analytics and data visualization
  • Optimized AWS Lambda functions for cost-effective serverless inference
  • Integrated OAuth2, JWT, and SAML for secure access control
  • Implemented monitoring with AWS CloudWatch and Dynatrace
PyTorch React.js AWS (S3, Lambda, SNS, SQS) LoRA RLHF Vision-LLMs OAuth2 JWT Docker CloudWatch Dynatrace
RadicalX

RadicalX GPT-4 Integration

University of North Texas | January 2023 - December 2024

Pioneered GPT-4 integration into the RadicalX educational platform, transforming the learning experience through advanced AI capabilities. The integration significantly enhanced platform interactivity and user engagement.

Key Achievements

  • Increased user engagement by 19% through enhanced platform interactivity
  • Seamlessly integrated GPT-4 API with existing platform architecture
  • Implemented responsible AI practices with focus on ethical deployment
  • Optimized API calls and response handling for improved performance

Technical Implementation

  • Integrated OpenAI GPT-4 API using LangChain for orchestration
  • Developed custom prompt engineering strategies for educational content
  • Implemented error handling and fallback mechanisms
  • Ensured compliance with responsible AI guidelines
  • Created monitoring and evaluation frameworks for AI system performance
GPT-4 OpenAI API LangChain Python Responsible AI API Integration
NLP Engine

NLP-Driven Recommendation Engine

University of North Texas | January 2023 - December 2024

Engineered and deployed NLP-driven recommendation engines that leverage advanced Natural Language Processing techniques to provide personalized learning recommendations across diverse educational modules.

Key Achievements

  • Achieved 25% increase in personalization accuracy
  • Deployed across diverse learning modules with consistent performance
  • Improved user learning outcomes through better content recommendations
  • Scaled to handle large volumes of user interaction data

Technical Implementation

  • Developed recommendation algorithms using collaborative and content-based filtering
  • Leveraged NLP techniques for semantic understanding of learning content
  • Fine-tuned transformer models for domain-specific recommendations
  • Implemented real-time recommendation generation with low latency
  • Created evaluation metrics for recommendation quality assessment
NLP PyTorch Hugging Face Transformers Recommendation Systems Machine Learning Python
Temenos

Temenos Infinity Spotlight

TEMENOS | March 2022 - December 2022

Played a key role in developing Temenos Infinity Spotlight, a browser-based console for data-driven configuration and seamless integration with core banking systems. Focused on creating intuitive UI/UX features with cross-browser compatibility.

Key Achievements

  • Reduced UI-related defects by 20%
  • Improved system stability through rigorous testing and bug fixes
  • Enhanced product usability through iterative feedback cycles
  • Reduced technical debt through code reviews and best practices

Technical Implementation

  • Developed responsive UI components with focus on user-centered design
  • Ensured cross-browser compatibility for banking applications
  • Implemented version control best practices using Git and Bitbucket
  • Participated in Agile/Scrum development processes
  • Collaborated with stakeholders to gather and implement requirements
JavaScript HTML/CSS Git Bitbucket Agile/Scrum UI/UX
ChatGPT

ChatGPT Integration & AI Assistant Platform

Lead Developer | 2024

Developed a comprehensive ChatGPT integration platform that leverages OpenAI's GPT models to create intelligent conversational AI assistants. Built end-to-end solutions for natural language processing, context management, and seamless API integration with advanced prompt engineering.

Key Features

Advanced Prompt Engineering

Implemented sophisticated prompt engineering techniques to optimize model responses, including few-shot learning, chain-of-thought prompting, and context-aware instructions.

Context Management

Built intelligent context management systems that maintain conversation history, manage token limits, and optimize context windows for better performance.

Multi-Modal Support

Integrated support for text, image, and code generation capabilities, enabling rich multi-modal interactions with the AI assistant.

Custom Fine-Tuning

Developed fine-tuning pipelines for domain-specific use cases, improving accuracy and relevance for specialized applications.

Technical Implementation

  • API Integration: Seamless integration with OpenAI GPT-3.5, GPT-4, and GPT-4 Turbo APIs with optimized request handling and error management
  • Backend Architecture: Built scalable Python/Django backend with async support for handling concurrent API requests and managing rate limits
  • Frontend Interface: Developed intuitive React.js chat interface with real-time message streaming, typing indicators, and conversation management
  • Prompt Engineering: Implemented advanced techniques including:
    • System prompts for role definition and behavior control
    • Few-shot learning with example-based prompting
    • Chain-of-thought reasoning for complex problem solving
    • Function calling for tool integration
  • Streaming Responses: Implemented Server-Sent Events (SSE) for real-time token streaming, providing instant feedback to users
  • Cost Optimization: Developed intelligent caching and token optimization strategies to reduce API costs while maintaining response quality
  • Error Handling: Robust error handling with retry logic, fallback mechanisms, and graceful degradation

Use Cases

  • Customer support chatbots with natural language understanding
  • Content generation and writing assistance tools
  • Code generation and programming assistance
  • Educational tutoring and Q&A systems
  • Data analysis and report generation
  • Creative writing and brainstorming assistants
OpenAI API GPT-4 Python Django React.js LangChain Prompt Engineering NLP SSE REST APIs WebSockets