ASAD
AHMAD
Building reliable backend systems, API platforms, and data-driven e-commerce workflows.
AI ENGINEER
ABOUT
ME
Computer Science Engineering student focused on backend services, API platforms, and dependable production workflows.
I build with Java, SQL, React, Next.js, Node.js, and Fastify. My work spans service reliability, marketplace-data analysis, REST APIs, CRUD workflows, Redis-backed systems, and practical AI security tooling.
SKILLS
Systems built, not just concepts studied. Click to deep-dive.
CORE CAPABILITIES
- LLM guardrails with real-time token inspection & policy enforcement
- Multi-vector prompt injection detection & mitigation
- PII/PHI masking with custom regex + ML classifier pipelines
- Vendor-agnostic LLM gateway orchestration (100+ endpoints)
- AI-powered phishing & anomaly detection (DistilBERT, 98%+ accuracy)
SYSTEMS BUILT
DEPTH
Sub-50ms real-time inference latency. Production-grade middleware with zero-trust architecture.
CORE CAPABILITIES
- Alpha signal generation with XGBoost, LSTM, and ensemble stacking
- SHAP-based model explainability for regulatory-grade transparency
- Asynchronous WebSocket order execution and position management
- High-frequency portfolio risk scoring with VaR and stress testing
- Spatial ML models (Random Forest) for geo-predictive intelligence
SYSTEMS BUILT
DEPTH
Low-latency async pipelines. Institutional-grade back-testing frameworks with live market feeds.
CORE CAPABILITIES
- Distributed Monorepo architecture (Next.js + Node.js) at scale
- RAG pipelines with pgvector, semantic search & intent routing
- Asynchronous job queues with BullMQ for 100+ concurrent webhooks
- SSR / ISR rendering strategies for SEO and performance optimization
- Row-Level Security (RLS) and Zod schema validation for data integrity
SYSTEMS BUILT
DEPTH
Production-grade monorepo at scale. 85%+ LLM accuracy on Hinglish dialect parsing in real-time.
CORE CAPABILITIES
- Containerized deployment pipelines using Docker + GitHub Actions CI/CD
- Serverless function architecture on Vercel and HuggingFace Spaces
- QLoRA fine-tuning of Qwen2.5 LLMs on cost-optimized cloud infra
- Redis-based caching layers for high-throughput message queues
- Cloud cost analysis with AWS/Azure cost API integration
SYSTEMS BUILT
DEPTH
Fine-tuned LLMs at minimal GPU cost. Automated right-sizing recommendations at infrastructure level.
CORE CAPABILITIES
- Geospatial analysis across 2000+ spatial cells for 9 urban zones
- Sustainability Digital Twin modeling (Carbon, Water, Energy KPIs)
- LP optimization engines for policy simulation & resource allocation
- Graph-based infrastructure network modeling (NetworkX)
- WHO-benchmarked urban planning with scenario impact simulation
SYSTEMS BUILT
DEPTH
70% R² score in urban density prediction. Real-time policy simulation affecting multi-billion public plans.
EXPERIENCE
E-COMMERCE TECHNOLOGY & OPERATIONS
JACOBDREAM LLCUsed Keepa API and SellerAmp SAS to analyze Amazon product data, historical pricing, sales-rank patterns, and profitability metrics for data-driven wholesale inventory decisions.
Evaluated margins, pricing volatility, and inventory risk across potential SKUs through structured sourcing and vendor-evaluation workflows.
WEB DEVELOPMENT INTERN
BIG FACTIONBuilt frontend modules, REST APIs, CRUD operations, and database-backed workflows for Margin Mate, an Amazon FBA analytics platform processing 50K+ e-commerce records.
Visit Big Faction ↗PROJECT SHOWCASE
ZENTRIS
Deployed service infrastructure with live health checks, structured logging, configuration validation, and resilient API controls.
SATURNX
Containerized MCP server that gives AI agents a safe, structured interface to authorized security workflows in isolated Kali environments.
TRADEVEX
Built ML-driven trading system with real-time WebSocket pipelines, SHAP-explainable signals, and automated risk controls.
MONEYLENSFIN
Full-stack wealth command center combining automated portfolio tracking, real-time risk engines (VaR, drawdowns), and AI-generated advisory insights.
INFRAVISION
Analyzed 2000+ spatial cells across Delhi NCR to forecast urban growth and infrastructure deficits with 70% R² prediction accuracy.
ZARIYA
Built cited-answer RAG engine over a classical knowledge base with expert-AI escalation when confidence drops below threshold.
REVORA
Eliminated revenue leakage for cloud kitchens by parsing Hinglish WhatsApp chats into structured orders via GPT-4o with 85%+ accuracy.
MINDGUARD
Achieved 98%+ phishing detection accuracy using DistilBERT with a hybrid heuristic validation layer for zero-day threat coverage.
OPTIOPS AI
Autonomous cloud cost optimizer using fine-tuned Qwen2.5 + RAG pipelines to generate actionable right-sizing recommendations from live infra data.
SHOPPILOT
Unified 10+ retail modules (CRM, Inventory, Analytics) into a single AI-driven commerce OS with WhatsApp storefront and zero-lag state sync.