Über
(RAG + RL + Behavioral Modeling)
Build 3 Core AI Models for a Fintech Revenue Optimization Platform
About the Project
We are building an AI-powered in-app engagement layer designed to increase user conversion, engagement, and revenue for fintech platforms.
Our system overlays intelligent tooltips, pop-ups, sidebars, and guided UI experiences directly inside web-based fintech applications (neobanks, lending platforms, crypto exchanges, trading platforms, etc.).
Project Scope
We are hiring a Machine Learning Engineer to build three production-grade AI models that power a real-time engagement engine:
Placement Model (Clustering model)
Content Generation Model (finetuned open source LLM + RAG)
Master Model (Reinforcement Learning model)
This is not a simple LLM wrapper project.
We are building a scalable AI infrastructure from the ground up.
Model 1: Placement Model
(Behavioral Clustering + Interaction Intelligence)
Objective:
Predict where to inject content in a fintech UI for maximum conversion impact.
Responsibilities:
Design a clustering model based on multiple fintech platform UIs
Convert DOM structure + clickstream + hover data into structured feature sets
Build unsupervised models (DBSCAN, Hierarchical Clustering, etc.)
Deploy as real-time API endpoint for placement decisions
Work with decision graphs and session replay / heatmap-style interaction data
Tech Stack:
Python
Scikit-learn
TensorFlow or PyTorch
FastAPI for serving
Model 2: Content Generation Model
(finetuned open source LLM + RAG)
Objective:
Generate the content of contextual, regulatory-aware fintech tooltips and in-app guidance in real time.
Responsibilities:
Implement Retrieval-Augmented Generation pipeline
Integrate vector database (Pinecone / Weaviate or equivalent)
Fine-tune open-source LLM (LLaMA / Mistral preferred)
Develop structured prompt templates for fintech tone & clarity
Implement evaluation metrics:
Build compliance validation layer (rule-based or hybrid)
Optimize inference cost and latency
Required Experience:
Production RAG systems
Vector DB architecture
Embeddings (Sentence-BERT or similar)
LangChain, Haystack, or equivalent orchestration framework
Content guardrails / filtering systems
Model 3: Master Model
(Reinforcement Learning model)
Objective:
Continuously learn the optimal intervention strategy to maximize:
Deposits
Loan completion
Feature adoption
Transaction volume
Responsibilities:
Design state-action-reward architecture
Define reward functions tied to conversion goals
Train using:
Deep Q-Learning OR
Policy Gradient methods
Orchestrate Placement + Content models
Handle streaming real-time feature inputs
Implement retraining pipelines & drift monitoring
Required:
Strong reinforcement learning experience
RLlib or equivalent frameworks
Experience with real-time decision systems
Understanding of exploration vs exploitation tradeoffs
Current Architecture
Backend: Python (FastAPI)
Database: MySQL
Frontend: Lightweight JavaScript widget injected via script tag
Modular microservices architecture
Cloud: AWS or GCP (GPU-enabled infrastructure)
Event logging & behavioral analytics pipeline
You will work directly with the founding technical team.
Important Constraints
Privacy-first architecture (no PII stored)
GDPR / SOC2 readiness
Low-latency real-time inference
Scalable to 100K–500K concurrent users
Cost-efficient GPU usage
Ideal Candidate
5+ years ML experience
Built at least one production RAG system
Built at least one reinforcement learning model beyond academic demos
Experience in fintech or regulated industries (strong plus)
Comfortable owning architectural decisions
Strong cost-awareness mindset
Engagement Details
Long-term contract (6+ months)
Part-time to start → potential expansion
Competitive rate (quality-focused)
Some US timezone overlap preferred
To Apply, Please Include:
Description of a production ML system you built
Example of a RAG pipeline you implemented
Example of an RL or behavioral optimization system
Your preferred ML stack
Your availability (hours/week)
Contract duration of 3 to 6 months. with 10 hours per week.
Mandatory skills: Python, Machine Learning, Data Science, AI Model Integration, AI Model Training
Sprachkenntnisse
- English
Hinweis für Nutzer
Dieses Stellenangebot stammt von einer Partnerplattform von TieTalent. Klicken Sie auf „Jetzt Bewerben“, um Ihre Bewerbung direkt auf deren Website einzureichen.