1. About Vibe

Vibe: Your AI Wingman. Built to help you with everything in your personal life.

2. Company Culture & Values

Initiative – Act before being asked

Ownership – Own work end to end

100x thinking – Big, bold ideas

Hard work – Clear hustler

Curiosity – Always digging deeper

3. The Role

You will architect and own Vibe's conversational AI, memory systems, voice infrastructure, and agent execution framework.

This means:

• Designing multi-LLM/MoE orchestration where specialized models handle emotion, reasoning, memory retrieval, task execution, and voice expression
• Building toward custom post-training (SFT/RLHF) with curated datasets for personality and emotional intelligence
• Owning the memory graph architecture — our core differentiator for continuity and personalization
• Creating real-time voice systems that feel like talking to a real person, not a bot
• Scaling all of this to production with sub-second latency

You're not integrating APIs. You're building the AI brain.

4. What You'll Own

Core AI Systems

• Design and implement multi-LLM orchestration for emotion, reasoning, memory, and task execution
• Build custom post-training pipelines (SFT/RLHF) for personality tuning and emotional intelligence
• Architect memory systems (short-term, long-term, relationship-level) — the foundation of human-like continuity
• Fine-tune and optimize models for natural, emotionally rich conversations that feel like texting your best friend

Voice & Real-Time

• Build real-time voice conversation systems with expressive, low-latency delivery
• Design and optimize streaming STT/TTS pipelines
• Handle real-time audio processing at scale

Infrastructure & Scale

• Build FastAPI services powering web + mobile
• Implement vector DB architectures for embeddings, retrieval, personalization
• Deploy on AWS (ECS Fargate, DynamoDB, CloudWatch, ECR, ALB)
• Use Docker for containerized deployments
• Ensure sub-second latency with thousands of concurrent users

5. About You

Experience

• 3–5+ years building AI/ML systems in production
• Proven track record of owning complex AI projects end-to-end

AI & Deep Learning

• Strong experience fine-tuning LLMs and building conversational agents
• Deep understanding of PyTorch and model optimization
• Experience with post-training techniques (SFT, RLHF, DPO) or strong desire to build this capability
• Understanding of transformer architectures, attention mechanisms, and model inference optimization
• Hands-on with LLM memory strategies, RAG, and retrieval-augmented personalization

Voice Systems

• Experience building real-time voice systems (STT/TTS pipelines)
• Knowledge of streaming audio processing and latency optimization

Backend & Infrastructure

• Expert-level Python + FastAPI
• Deep experience with vector databases (Pinecone, Weaviate, Qdrant) for production workloads
• Strong understanding of Redis for caching and performance
• Comfortable with Docker and deploying to AWS (ECS, DynamoDB, load balancing)
• Experience scaling AI systems for production traffic

Bonus Points

• Experience with multi-agent systems or mixture-of-experts architectures
• Built custom training pipelines or worked with training data at scale
• Background in dialogue systems or emotional AI
• Open-source contributions to AI/ML projects

6. What Success Looks Like

• Feels Human: Conversations are warm, natural, emotionally aware — users forget they're talking to AI
• Remembers Everything: Vibe builds real continuity and adapts over time — memory is our superpower
• Voice Feels Real: Real-time calls with smooth, expressive responses that feel like talking to a friend
• Handles Life: Agent reliably executes tasks, follows up, and helps in daily life
• Scales Gracefully: Fast, stable, sub-second latency at thousands of concurrent users
• Users should say: "Talking to Vibe changed my life"