We are developing a real-time voice AI platform that combines speech recognition, LLM-driven reasoning, and natural-sounding text-to-speech capabilities. This is production-grade software designed to run continuously, scale under heavy load, and deliver human-like conversations with minimal latency.
If you are a seasoned Python engineer who thrives on building resilient systems and solving tough concurrency challenges, this is the role for you.
What You’ll Do/Essential Job Functions
● Architect and build the conversation orchestration service: ASR → LLM inference → TTS streaming in real time
● Write robust, asynchronous Python code designed to handle high concurrency without deadlocks, race conditions, or memory leaks
● Design and maintain clean, well-structured APIs for future scalability and ease of debugging
● Manage interaction data using SQLAlchemy (or equivalent) with efficient schema design and safe migrations
● Implement observability: structured logging, metrics, and tracing across the system for instant issue diagnosis
● Partner with ML and Product teams to rapidly iterate on conversation flow and user experience
● Enforce a strong testing culture: automated unit tests, E2E flows, and load testing
● Build resilient systems capable of handling real-world edge cases like noisy audio, unreliable APIs, and flaky networks
● Continuously profile, optimize, and reduce latency and response times
Requirements
What We Expect You To Know/Requirement
● Deep Python expertise: 5+ years in Python, production systems experience required, context managers, generators, event loops, GIL, and effective use of asyncio
● Database fundamentals: data modeling, efficient queries, ORM best practices
● Networking & I/O: streaming, backpressure, and resilient design for unreliable networks
● Testing discipline: delivering production-ready, validated code
● Observability mindset: metrics, logs, and traces are integral to your coding process
● Production readiness: You’ve built and supported systems running live at scale.
What You’re Like
● Curious: You don’t just fix bugs — you find root causes
● Calm under pressure: You can diagnose incidents, resolve them quickly, and prevent recurrences
● Pragmatic: You solve problems without over-engineering
● Collaborative: You write code for your teammates and your future self
● Quality-driven: You refuse to compromise on correctness and reliability
● Data-informed: You make decisions based on real latency metrics, throughput, and error rates
What To Expect
This is not a feature-factory role. You will be responsible for building a real-time system that stays online, hits latency targets, and performs reliably under pressure.
Team Culture
● Direct, collaborative, and low on politics
● High ownership: see your work running in production, serving real users
● Obsessed with clarity, correctness, and reliability
● Fast-moving: minimal ceremony, maximum impact
● Pragmatic planning: no endless poker sessions — scope, assign, design, deploy.
Your Daily Work
● Ensure performance, reliability, and observability in everything you build
● Collaborate closely with ML and product teams on speech recognition, TTS voices, and LLM behavior.
● Monitor, debug, and improve the system as it runs in production.
Working Terms: The candidate must be flexible and work during US hours at least until 6 PM ET, which is essential for this role.
Top Skills
What We Do
Checkmate empowers enterprise restaurant brands with powerful ordering solutions and hands-on support. Our scalable technology enables restaurants to drive sales across channels, including custom websites, apps, kiosks, catering, third-party marketplaces, voice AI, and more. With seamless integrations, smarter analytics, and 24/7 service, Checkmate helps brands conquer their digital goals. Restaurants can launch unique ordering experiences, centrally manage menus, recapture revenue, leverage customer data, and continually adapt with new integrations. Regardless of how you want to grow, Checkmate has the tools and guidance to power, manage, and evolve your digital business.
.png)
