> SYSTEM 1.0 ONLINE ...

FRANCIS

_
View Projects Download CV
SCROLL

// ABOUT_ME

Francis - AI Engineer
Francis AI Engineer & Full-Stack Developer
A.I... / protocol_alpha

$ whoami

AI Engineer


$ cat bio.txt

4 years of backend engineering in C#/.NET and 2+ years

building production-grade LLM systems — multi-model

routing, RAG pipelines, ReAct agents, and MCP on AWS.

Proven focus on latency optimization and cost efficiency.


$ cat focus.txt

LangChain · LangGraph · CrewAI · MCP · AWS


$ ▋

0 Projects Built
0 AI Models Integrated
0 AWS Services Used
0 % Open Source

// TECH_STACK

Programming Languages

Python C++ C#

Backend & API

FastAPI REST APIs WebSockets .NET

LLM Systems

RAG ReAct Agents Multi-Agent Systems Prompt Engineering LangChain LangGraph CrewAI MCP

Databases

MySQL FAISS Pinecone Chroma OpenAI Embeddings Sentence Transformers

Cloud & DevOps

AWS ECS Fargate EC2 App Runner S3 Bedrock AgentCore Docker CI/CD GitHub Actions

// EXPERIENCE

Capgemini

Associate I Bengaluru, Karnataka, India
February 2022 – Present
  • Engineered a content extraction pipeline using C#/.NET, reducing manual processing effort by ~60% across enterprise workflows.
  • Optimized SQL Server queries and data pipelines, improving throughput for large unstructured datasets.
  • Built scalable backend systems for enterprise workflows handling unstructured data at scale.
  • Delivered features across full SDLC — design, development, deployment, and performance optimization.
  • Developed an Intelligent Query Routing Platform integrating multiple LLM providers (OpenAI, Groq, Gemini, Llama) with dynamic fallback strategies based on cost and latency; delivered p95 500–900ms.
  • Architected an Enterprise Knowledge Retrieval System using LangChain + LangGraph with ReAct-based tool orchestration over 1K–3K document chunks via FAISS; integrated AWS Bedrock AgentCore for persistent conversational memory, achieving ~85% retrieval accuracy (hit@5) at p95 ~1.2s.

// PROJECTS

01
👁️ 🔊 🎨 🔍
AI

Intelligent Query Routing Platform

Multi-Model LLM Routing System

Designed a multi-provider LLM routing layer integrating OpenAI, Groq, Gemini, and Llama with dynamic fallback and routing strategies based on response time and cost constraints. Implements tool calling with web search for real-time knowledge augmentation across text, vision, TTS, and image generation.

FastAPI Groq Gemini Stable Diffusion OpenAI
p95 ~500ms ~$0.0002/req
02
Q → VEC → GEN

Enterprise Knowledge Retrieval System

Retrieval-Augmented Generation

Built a document Q&A system using LangChain + FAISS with semantic retrieval over 1K–3K document chunks. Designed a ReAct workflow using LangGraph with Tavily fallback, reducing failed queries by ~30–40%; integrated AWS Bedrock AgentCore for persistent memory. Attained ~85% retrieval accuracy at p95 ~1.2s.

LangChain LangGraph FAISS OpenAI Tavily
p95 ~1.2s hit@5 ~0.85
03

AI Code Review Crew

Multi-Agent Pipeline

5-agent sequential pipeline (CrewAI) specialized in bug detection, OWASP Top 10 security, performance optimization, and documentation analysis. Processes up to 50 files/run with severity scoring, cutting code review turnaround by ~40%.

CrewAI FastAPI AWS ECS OpenAI Anthropic
~60s/file ~$0.008/review
View Code
04
Q A 📝

Study Tools MCP

MCP-Based AI Assistant

Claude Desktop-integrated study assistant using Model Context Protocol for structured tool interaction with document parsing pipelines (PDF, Markdown). Generates quizzes, flashcards, summaries & explanations — 100+ pages/session.

MCP FastAPI OpenAI PyPDF2 AWS EC2
p95 ~800ms ~$0.0005/call
View Code
05
? 💬 ?

Lauki FAQ Agent

Conversational Support AI

Stateful conversational AI agent for FAQ-based customer support with persistent memory via AWS Bedrock AgentCore. Semantic search using FAISS + sentence-transformers over ~500–1K FAQ entries; reduced cold-start latency via S3-backed vector store.

LangGraph AWS Bedrock FAISS Groq Docker
p95 ~600ms accuracy ~0.87
View Code
06

Expense Manager

Full-Stack Finance App

Full-stack expense tracker with glassmorphism UI, category-based analytics, real-time charts, and 41+ automated test cases. Deployed on AWS Elastic Beanstalk with MySQL RDS.

FastAPI MySQL AWS Beanstalk Pydantic JavaScript
View Code

> ESTABLISH_SECURE_LINK

Let's Build Something

Open to collaborations, opportunities, and interesting projects in AI engineering.

GitHub LinkedIn Email