PARIS // 48.8566° N, 2.3522° E

ANANDHU KRISHNA

vLLM • llama.cpp • GCP
Agentic RAG • Neo4j

AI Engineer & Data Scientist architecting intelligent systems at the intersection of Large Language Models, autonomous agentic workflows, and scalable machine learning. Transforming complex research prototypes into high-performance, enterprise-grade production.

Explore Works Curriculum Vitae
01 // Neural Profile

Curating Knowledge.
Zero Syntax Errors.

Anandhu Krishna
Anandhu Krishna
Paris, France
Active

An AI engineer on a mission to explore the world and tick off a carefully curated bucket list executing goals without hitting any syntax errors.

Master's graduate from EPITA Paris in Computer Science (Data Science & Analysis), with hands-on experience designing and operating production AI systems at Hubicus (IPSOS), AXA Group Operations, and 6D Technologies.

Location
Paris, France
Specialization
LLMs & Agentic AI
Master's Degree
EPITA Paris (Bac +5)
Languages
English • French • Malayalam
02 // Trajectory & Research

Engineering Trajectory

Ipsos BVA Logo

AI Engineer

Hubicus (Ipsos BVA) • Paris, France
06/2026 — Present
Local Model Deployment & Serving: Architecting sovereign on-premise and private cloud inference engines using vLLM and llama.cpp with AWQ and GGUF quantization for ultra-low latency and maximum token throughput.
GCP Model Garden & Vertex AI: Deploying, fine-tuning, and benchmarking foundational open-weight models via Google Cloud Platform Model Garden to power high-scale enterprise workflows.
Inference Optimization & Latency Reduction: Implementing continuous batching, PagedAttention, KV cache management, and speculative decoding to slash compute costs and deliver sub-second TTFT SLAs.
Production Sovereign AI Workflows: Delivering enterprise-grade LLM runtimes integrated into customer intelligence products with zero data leakage, audit logging, and EU AI Act compliance.
Ipsos BVA Logo

AI Engineer Intern

Hubicus (Ipsos BVA, Customer Insights) • Paris, France
11/2025 — 04/2026
AI Services Integration: Integrated multimodal AI evaluations across customer channels (chat, email, audio) into the flagship quality monitoring SaaS platform.
LLM-Agnostic Architecture: Decoupled vendor locks to allow dynamic model routing across Mistral, OpenAI, and Gemini models.
EU AI Act & Security: Implemented PII redaction pipelines, audit telemetry, and prompt-injection guardrails complying with the European AI Act.
MCP Server & RAG Agent: Developed a Model Context Protocol server exposing platform APIs and an internal knowledge retrieval agent.
AXA Logo

AI/ML Engineer Intern

AXA Group Operations • Paris, France
03/2025 — 09/2025
Computable Contracts Engine: Built the core ingestion entry point parsing intricate insurance contracts into structured semantic graphs.
25+ Parser Benchmark & GPT-4o: Benchmarked over 25 document parsing architectures; adopted multimodal GPT-4o with custom scoring (NID, TEDS).
Graph Intelligence: Modeled contractual obligations and entities using Kuzu and Neo4j graph databases for high-speed multi-hop queries.
Production Observability: Introduced Langfuse for tracing LLM execution, token economics, and latency in Dockerized CI/CD workflows.
6D

Data Scientist

6D Technologies • Bangalore, India
02/2021 — 08/2023
Customer Churn Reduction: Engineered predictive retention pipeline reducing customer turnover by 15% for international telecom carriers.
Upsell & Cross-Sell Systems: Raised customer engagement by 25% through market-basket analysis (Apriori, FP-Growth, Matrix Factorization).
Forecasting & Similarity: Built XGBoost time-series pipeline predicting next-purchase intervals with 62% accuracy across millions of records.
EPITA

MSc Computer Science — Data Science & Analysis

EPITA - School of Engineering and Computer Science • Paris, France (Bac +5)
2023 — 2025

Deep theoretical and applied curriculum encompassing neural network architectures, distributed data engineering, cloud platforms, and statistical modeling.

03 // Spatial Exhibition

Selected Engineering & Research

Automated Modeling of Insurance Contracts & Knowledge Graphs
AXA Group Operations • Enterprise AI
PROJECT // 01

Computable Contracts via Multimodal LLMs & Graphs

Production-ready entry module for AXA's Computable Contracts platform. Evaluated 25+ document parsers and engineered a GPT-4o multimodal parsing pipeline with Neo4j/Kuzu graph database integration, Clean Architecture, and Langfuse observability.

GPT-4o Neo4j Kuzu Graph Clean Architecture Langfuse Docker
View Repository
Agentic RAG & Model Context Protocol System
Hubicus • Autonomous Agents
PROJECT // 02

MCP Server & RAG Knowledge Agent

Engineered an enterprise Model Context Protocol (MCP) server exposing SaaS platform endpoints and built an autonomous RAG-driven knowledge retrieval agent. Features prompt injection guardrails, PII redaction, and EU AI Act compliance.

MCP Protocol RAG FastAPI Vector DB EU AI Act Guardrails
View Repository
Zero-Shot LLMs & Attention Representation Research
Research • Deep Learning
PROJECT // 03

Enhancing Zero-Shot Capabilities in Large Language Models

In-depth research and experimental evaluation analyzing prompt heuristics, parameter-efficient fine-tuning (PEFT), and attention representations to maximize zero-shot generalization across domain-specific NLP benchmarks.

Transformers PyTorch Zero-Shot Hugging Face Fine-Tuning
View Research Repo
Telecom Predictive Churn & Recommendation Engine
6D Technologies • Big Data
PROJECT // 04

Predictive Churn & Upsell Recommendation Engine

Architected telecom machine learning pipeline cutting churn by 15% and boosting engagement by 25%. Employed Matrix Factorization, FP-Growth, Apriori market-basket analysis, and XGBoost temporal purchase forecasting.

XGBoost Matrix Factorization FP-Growth Time Series Big Data
Case Study Details
04 // Neural Constellation

Technical Capabilities

<AI> LLMs & Deep Learning
LLMs vLLM llama.cpp Gemini Mistral GPT-4o Prompt Engineering MCP Protocol Hugging Face Transformers PyTorch Langfuse RAG
<OPS> MLOps & Engineering
Local Model Deployment Docker FastAPI Git CI/CD Airflow Streamlit Clean Architecture Poetry Linux RabbitMQ
<GRAPH> Graph Databases & Data
Neo4j Kuzu Pandas NumPy SciPy Dask SQL BigQuery
<CLOUD> Cloud & Infrastructure
GCP Model Garden Vertex AI Google Cloud (GCP) AWS Cloud Practitioner Microsoft Azure (AZ-900) Firestore
05 // Direct Uplink

Initiate Transmission

Let's Build Intelligent Realities.

Open to high-impact AI Engineering opportunities, agentic architecture research, and technical consultancy in Paris and globally.

anandhukrishna091@gmail.com +33 07 80 86 43 13
Paris, France
[ENCRYPTION: TLS 1.3] • [LATENCY: ~12ms]