Job Details

At ASAPP, our mission is simple: deliver the best AI-powered customer experience-faster than anyone else. We are guided by principles that shape how we think, build, and execute, including deep customer obsession, purposeful speed, ownership, and a relentless focus on outcomes. We work in small, highly skilled teams, prioritize clarity over complexity, and continuously evolve through curiosity, data, and craftsmanship.

We're building a globally diverse team of technologists and problem solvers who thrive in fast-paced environments, value collaboration, and approach every challenge with a Day 1 mindset. With hubs in New York City, Mountain View, Latin America, and India. If you're driven by continuous learning, rapid iteration, and the challenge of building in a high-growth startup, this is more than a role-it's a journey.

We are seeking a Senior Speech Software Engineer to drive both the infrastructure and applied speech intelligence behind our real-time voice AI platform. This is not just a systems role - you will operate at the intersection of speech research, model optimization, and production engineering, ensuring our ASR and TTS systems meet the demanding quality, latency, and reliability requirements of enterprise call centers.

You will help evolve our speech stack to deliver human-like, low-latency voice interactions at massive scale, tuning and adapting modern speech models to perform in noisy, real-world customer environments. You will work closely with Speech Scientists, ML Researchers, and Infrastructure Engineers to bridge cutting-edge speech technology with hardened production systems.

What you'll do

• Speech Model Optimization & Applied Research:

Tune and optimize ASR and TTS models for real-world call center environments, improving transcription accuracy, noise robustness, and speaker variability

Improve spoken output naturalness by refining prosody, pacing, number and spelling pronunciation, and conversational flow

Balance latency vs. quality tradeoffs in streaming speech pipelines to maintain real-time responsiveness

Evaluate and integrate emerging speech technologies (e.g., noise suppression, voice activity detection, diarization) to measurably improve performance

• Voice Infrastructure & Systems Engineering

Architect and modernize a scalable, high-availability voice infrastructure that replaces legacy systems

Build multi-threaded, low-latency server frameworks capable of handling thousands of concurrent real-time audio streams

Design and operate streaming ASR - LLM - TTS pipelines that power live AI-driven customer conversations

Develop robust media stream handling to ensure reliable audio flow between telephony providers, clients, and ML services

• Evaluation, Observability & Quality

Define and implement speech quality evaluation frameworks, including WER/CER analysis, latency tracking, and perceptual TTS metrics

Build tooling and dashboards to monitor production performance and detect regressions in accuracy, latency, or naturalness

Create load-testing and simulation tools to model high-concurrency, real-world voice traffic

• Cross-Functional Collaboration

Partner with Speech Scientists and ML Researchers to productionize new ASR and TTS models

Work with Security and Compliance teams to ensure voice data handling meets enterprise and regulatory standards

Collaborate with Product teams to translate conversational quality requirements into measurable system improvements
>

What you'll need

• Core Engineering Background

5+ years of software engineering experience building and operating production-grade distributed systems

Strong proficiency in Golang or Python (or willingness to become an expert quickly)

Experience designing low-latency, high-concurrency systems, ideally involving real-time media or streaming data

• Speech & Audio Expertise

Practical experience working with ASR and/or TTS systems in applied or production environments

Understanding of how to adapt and tune speech models for domain-specific use cases

Familiarity with speech quality metrics such as WER, CER, MOS, latency, and streaming stability

Strong grasp of audio fundamentals, including sample rates, codecs (Opus, G.711), buffering, packet loss, and jitter

• Applied ML for Speech

Experience evaluating model performance and running structured experiments to improve transcription accuracy and speech naturalness

Comfort working with modern ML tooling and model APIs to fine-tune, adapt, or post-process speech model outputs

Ability to make pragmatic tradeoffs between model quality, compute cost, and real-time constraints
>

What we'd like to see

• Experience with noise reduction, echo cancellation, VAD, diarization, or other speech enhancement technologies
• Familiarity with forced alignment techniques or phoneme/word-level timing models
• Hands-on experience deploying ML services with Kubernetes, Docker, and cloud platforms (AWS/GCP/Azure)
• Knowledge of event-driven and asynchronous systems (e.g., async I/O, event loops, streaming frameworks)
• Experience analyzing large-scale speech or conversation datasets to drive model or system improvements

$215,000 - $235,000 a year

The compensation includes salary plus performance bonus. The actual salary may be different depending upon non-discriminatory factors such as qualifications, experience, and other factors permitted by law.

ASAPP is committed to creating a diverse environment and is proud to be an equal opportunity employer. All qualified applicants will receive consideration for employment without regard to race, color, religion, gender, gender identity or expression, sexual orientation, national origin, disability, age, or veteran status. If you have a disability and need assistance with our employment application process, please email us at [email protected] to obtain assistance. #LI-AG1 #LI-Hybrid

We may use artificial intelligence (AI) tools to support parts of the hiring process, such as reviewing applications, analyzing resumes, or assessing responses. These tools assist our recruitment team but do not replace human judgment. Final hiring decisions are ultimately made by humans. If you would like more information about how your data is processed, please contact us.