| Project | Domain & Architecture | Core Stack |
|---|---|---|
| ποΈ voice-ai-low-latency-agent | Real-Time Low-Latency Voice AI Agent & Speech RAG Pipeline (Sub-380ms TTFA) | Python WebRTC Whisper LoRA (STT) Streaming TTS Qdrant |
| π₯ clinical-ai-medical-rag-assistant | Clinical AI Scribe, Multimodal Medical Imaging Diagnosis & HL7 FHIR R4 / SOAP System | Python Medical VLM FHIR R4 ICD-10 NER Safety Guardrails |
| π agentic-rag-langgraph-pipeline | Stateful Corrective & Hybrid Agentic RAG with Reciprocal Rank Fusion (RRF) | LangGraph Qdrant Hybrid RRF Claude Sonnet FastAPI |
| βοΈ multi-agent-mcp-orchestrator | Multi-Agent LLM Orchestration Platform with Model Context Protocol (MCP) Server Bus | MCP Multi-Agent Anthropic Tool-Use JSON-RPC 2.0 Docker |
| β‘ llm-finetuning-eval-pipeline | 4-Bit QLoRA / LoRA SFT Post-Training Pipeline & Automated CI/CD Quality Gate | PEFT / LoRA TRL SFTTrainer P01-P12 Diagnostics GitHub Actions |
π Building Production AI Agent Harnesses β’ Native LLM Tool-Calling Pipelines β’ Scalable Full-Stack Systems
I am a software engineer specializing in Autonomous AI Agents, Generative AI systems, and full-stack architecture. I build robust, production-grade agentic harnesses from first principlesβcombining multi-turn ReAct reasoning loops, dynamic tool reflection, and durable memory layers with high-performance Python and MERN backends.
- π Current Focus: Autonomous AI Receptionists, Multi-Agent Coordination, and Scalable Full-Stack SaaS Platforms.
- π§ Specialized In: ReAct Agent Loops, LLM Function Calling, Dynamic Schema Generation, and Durable Memory Stores.
- β‘ Core Engineering Values: First-principles design over framework bloat, defensive error handling, and 100% deterministic testability.
- π¬ Ask Me About: AI Agent Architectures, Python / FastAPI, MERN Stack (React, Node, Express, MongoDB), and REST APIs.
| Category | Technologies & Tools |
|---|---|
| AI & Agentic Systems | Google Gemini API β’ OpenAI GPT-4o β’ Ollama β’ ReAct Agent Loops β’ Function Calling β’ Prompt Engineering |
| Backend Engineering | Python β’ FastAPI β’ Node.js β’ Express.js β’ REST APIs β’ Pydantic β’ httpx β’ WebSockets |
| Frontend Development | React.js β’ JavaScript (ES6+) β’ TypeScript β’ HTML5 / CSS3 β’ Tailwind CSS β’ Redux |
| Databases & ORM | PostgreSQL β’ MongoDB β’ Redis β’ SQLite β’ Mongoose β’ SQLAlchemy |
| DevOps & Testing | Docker β’ Docker Compose β’ Git / GitHub β’ pytest β’ CI/CD Pipelines β’ Linux |
Production-Grade AI Receptionist built from First Principles (Zero Framework Bloat)
An autonomous AI receptionist capable of conversational table bookings, real-time availability checks,
dietary filtering, food pre-ordering, and personalized recommendations using durable customer memory.
- π Autonomous ReAct Loop: Multi-turn reasoning engine that safely chains availability checks, table bookings, and food orders in sequence.
- π οΈ 16 Specialized Tools: Python reflection decorator (
@tool) that dynamically generates schemas from type hints with defensive fault recovery. - π§ Durable Memory Engine: Persists customer preferences (seating, allergies, favorites) and past order history into the backend database.
- π§ͺ 100% Deterministic Testing: 56 unit tests utilizing
StubLLMandMockAPIfixtures running in<0.4swith zero real LLM costs.
β‘ AI SaaS SoftwareA full-stack AI SaaS application featuring modern UI components, subscription billing workflows, and generative AI integrations for automated content workflows.
|
π¦οΈ Weather AppReal-time meteorological web application consuming dynamic weather APIs to deliver accurate climate forecasts, temperature metrics, and geolocation lookups.
|
π Password GeneratorCryptographic utility for generating highly secure, customizable passwords with entropy controls (symbols, digits, casing, length) and one-click clipboard copy.
|
βοΈ Full-Stack RepositoriesA collection of web applications, REST backend services, and algorithmic problem-solving implementations built with clean code and modern standards.
|
