Founding Backend & AI Systems Engineer

About AMOGHA AI

At Amogha AI, we are building Jamun, India's first voice-led conversational AI app specifically for mental health, therapy, and emotional well-being. Our mission is to provide a supportive, empathetic listener in your pocket, 24/7. We are creating a next-generation product that understands user context, provides therapy-grade support, and guarantees 100% safety and privacya truly specialized application far beyond a generic ChatGPT experience.

THE OPPORTUNITY

We are nearing completion of the MVP of Jamun. All core AI modules (ASR, LLMs, TTS, RAG, Memory) are functional.

Founding Backend & AI Systems Engineer

And, when applying, besides the CV, highlight relevant experiences/skillset against the expected responsibilities.

Full-time / Significant part-time | Remote/Hybrid | Competitive compensation | 0.5% - 2% Equity

WHAT YOU'LL DO

1.Orchestration & Advanced State Management

• Persistent Session State:

  Architect a robust state management system that maps user sessions to specific nodes in our multi-layered prompt logic.

• Prompt Injection:

  Upon resumption, the orchestration layer dynamically injects the correct Context Window and System Prompt Layer into the LLM so it knows exactly where it left off without the user having to remind it.

2.Latency Optimisation (The Speed Layer)

• Obsess over Time to First Byte (TTFB) and end-to-end voice latency.

• Streaming architecture:

  Replace request-response cycles with full-duplex streaming (WebSockets). Ensure TTS starts playing audio while the LLM is still generating text.

• Asynchronous Processing:

  Refactor the codebase to ensure heavy tasks (like vector database lookups) never block the LLM thinking/processing thread.

3.Production Infrastructure (The Scale Layer)

• Concurrency Handling:

  Ensure the server infrastructure can handle 10-50 simultaneous active sessions. This involves optimising socket connections and thread/process management.

• Modular Deployment:

  Containerise every component separately so we can swap/upgrade individual modules without downtime. Set up the initial CI/CD pipeline for smooth production updates.

• API Gateway/Load Balancing:

  Set up the entry point for the app to manage traffic efficiently.

4.Reliability & Error Recovery (The Stability Layer)

• Graceful Degradation:

  Build logic for what happens when X fails. (e.g., if RAG times out, fall back to base logic).

• State Management:

  Ensure user session state (context) is preserved even if a connection drops and reconnects.

• Observability:

  Implement logging and monitoring (e.g., Sentry, Prometheus) to track latency per module and catch errors in real-time.

WHO YOU ARE

You'll thrive in this role if you are:

• Obsessed with social content you immediately spot trends and know how to adapt them
• Strong at storytelling, with the ability to turn any idea into a scroll-stopping post
• Comfortable with design tools like Canva (bonus for CapCut, Figma, Premiere Rush)
• Comfortable working independently in a fast-paced environment
• Curious about AI and excited to simplify technical ideas for everyday users

WHAT WE'RE LOOKING FOR

1.

• 5+ Years Experience:

  Proven track record of taking Python-based MVPs to production.

• Async Proficiency:

  Expert-level knowledge of asyncio (Python) or Goroutines (Go). You understand the difference between CPU-bound and I/O-bound tasks and how to optimise for both.

• WebSockets & Streaming:

  Deep experience with WebSockets, Socket.io, or gRPC. You have built apps where the server pushes data to the client in real-time (not just REST APIs).

2.

• Finite State Machines (FSM):

  Experience implementing state machines in code (e.g., using transitions library in Python) to track user progress through complex logic flows.

• LLM Orchestration:

  Experience employing frameworks like LangChain or LlamaIndex in production. You understand prompt chaining and state management.

• Database Interactions:

  Proficiency with efficient database querying (PostgreSQL, Redis) and interacting with Vector Databases (Pinecone/Milvus).

• Queue Management:

  Experience with Redis/RabbitMQ/Celery for background tasks (like vectorising memories) so they don't slow down the active conversation.

• API Design:

  Experience designing clean, internal APIs for modular communication.

3.

• Debugging:

  You are comfortable debugging complex distributed systems where an error might be caused by a timeout in one service affecting another.

• Load Testing:

  Experience with tools like Locust or K6. You shouldn't just hope the system handles 50 users; you must simulate 100 users to break it, find the bottlenecks, and fix them.

4.

• Voice AI Experience:

  You have worked with audio buffers, sampling rates, and chunking for ASR/TTS services previously.

• Privacy & Security Engineering:

  Familiarity with PII redaction (using tools like Presidio/Spacy) and implementing basic encryption standards (AES/KMS) to secure sensitive user data.

AND startup GRIT!

We believe interviews alone can't assess fit on either side. Our process centres around paid work trials, which can be done remotely. You will work with us on real problems, collaborate as peers, and get a genuine sense of what working together would feel like.

The bar is high, but the reward is immense!