Hiring Speech Engineers: The Complete Guide

# Speech Engineer

**Location:** Seattle, WA (Hybrid) · **Employment Type:** Full-time · **Level:** Mid-Senior

## About [Company]

[Company] is building the next generation of voice AI for enterprise communication. Our speech technology powers 50 million+ business calls monthly, transforming voice data into actionable insights for sales, support, and compliance teams.

We process audio in real-time with <500ms latency, achieving 95%+ accuracy on business conversations across multiple domains. Our speech team is core to the product value we deliver.

**Why join [Company]?**

- Work on speech at meaningful scale—50M+ monthly calls, real-time processing
- Solve hard speech problems (domain adaptation, speaker diarization, noise handling)
- Direct product impact—accuracy improvements translate to customer value
- Join a 10-person speech team within a 100-person engineering org
- Series C company with strong enterprise traction
- Competitive compensation with meaningful equity

## The Role

We're looking for a Speech Engineer to help us build world-class ASR and voice analysis systems. You'll work on speech recognition, speaker diarization, and the voice AI features that make our product valuable.

This is core speech engineering. You'll improve accuracy on challenging audio (phone calls, multiple speakers, domain vocabulary), optimize for real-time streaming, and build the evaluation systems that measure progress. The LLM wave has accelerated speech technology—you'll work with cutting-edge approaches.

The ideal candidate has deep speech experience and understands the unique challenges: acoustic modeling, language modeling, and the gap between research benchmarks and production performance.

## What You'll Work On

### First 30 Days
- Onboard to our speech stack, ASR models, and evaluation systems
- Understand our audio domains, accuracy challenges, and customer needs
- Ship your first improvement (model tweak, feature, or evaluation enhancement)

### First 90 Days
- Own a speech system component (ASR, diarization, or voice activity detection)
- Contribute to model improvements with measurable accuracy gains
- Build or improve evaluation pipelines
- Collaborate with product on voice AI roadmap

### First Year
- Lead speech initiatives with direct customer impact
- Architect improvements to our speech infrastructure
- Develop new capabilities (additional languages, domains, or features)
- Mentor teammates on speech engineering best practices

## Objectives of This Role

- Improve ASR accuracy (WER reduction) on business conversations
- Reduce real-time processing latency while maintaining quality
- Improve speaker diarization accuracy for multi-speaker calls
- Build robust evaluation systems that catch regressions
- Expand domain coverage for industry-specific vocabulary

## Responsibilities

**ASR Development (50%)**
- Build and improve ASR models for business speech domains
- Fine-tune and adapt models for domain-specific vocabulary
- Optimize models for real-time streaming (low latency, high accuracy)
- Implement language model improvements
- Handle acoustic challenges (phone audio, background noise, accents)

**Voice AI Features (30%)**
- Develop speaker diarization and identification systems
- Build voice activity detection and audio segmentation
- Implement speaker embedding and voice matching
- Create custom vocabulary and terminology handling
- Build audio preprocessing and enhancement pipelines

**Evaluation & Infrastructure (20%)**
- Build evaluation frameworks to measure accuracy and regressions
- Create test sets and benchmarking infrastructure
- Monitor production accuracy and investigate issues
- Optimize serving infrastructure for latency and cost
- Collaborate with ML platform on training infrastructure

## Required Skills and Qualifications

**Technical Skills**
- 5+ years of ML engineering experience
- 3+ years specifically in speech/audio processing
- Deep experience with modern ASR architectures (transformer-based, CTC, attention)
- Strong PyTorch expertise for model development
- Real-time streaming system experience
- Understanding of audio signal processing fundamentals

**Speech Knowledge**
- Familiarity with speech corpora and evaluation metrics (WER, CER)
- Experience with acoustic modeling and language modeling
- Understanding of audio preprocessing (feature extraction, noise handling)

**Soft Skills**
- Analytical approach to debugging accuracy issues
- Clear communication about model tradeoffs
- Collaboration with product and data teams

## Preferred Skills and Qualifications

- Multi-lingual ASR experience
- TTS (text-to-speech) development experience
- Telephony audio experience (8kHz, compression artifacts)
- Experience with Whisper, Wav2Vec, Conformer, or similar
- Publications at INTERSPEECH, ICASSP, or similar
- Background in linguistics or phonetics

## Tech Stack

**ML:** PyTorch, Hugging Face Transformers, custom speech models

**Speech Models:** Whisper, Wav2Vec, Conformer, custom fine-tuned models

**Audio:** FFmpeg, librosa, custom audio processing

**Streaming:** Apache Kafka, custom real-time pipelines, WebSocket

**Infrastructure:** AWS (SageMaker, EC2 with GPUs), Kubernetes

**Evaluation:** Custom WER evaluation, internal benchmarks

## Compensation and Benefits

**Salary:** $160,000 - $210,000 (based on experience)

**Equity:** 0.05% - 0.15% (4-year vest, 1-year cliff)

**Benefits:**

*Health & Wellness*
- Medical, dental, and vision insurance (100% covered)
- $100/month wellness stipend
- Mental health support through Lyra

*Time Off*
- Unlimited PTO with 15-day minimum encouraged
- 11 paid company holidays
- 16 weeks paid parental leave

*Professional Development*
- $3,000 annual learning budget
- Conference attendance (INTERSPEECH, ICASSP)
- Internal ML reading groups and tech talks

*Financial*
- 401(k) with 4% company match
- Equity refresh grants annually

*Workspace*
- $1,500 home office setup allowance
- Hybrid work (2-3 days in Seattle office)
- GPU access for local development

## Engineering Culture

**How we work:**
- Accuracy obsession—every percentage point matters
- Rigorous evaluation before production deployment
- Close collaboration between speech and product teams
- Regular model reviews and technical deep dives
- Blameless post-mortems after issues

**What we value:**
- Production accuracy over benchmark performance
- Real-time reliability for customer-facing systems
- Domain expertise in speech and audio
- Continuous learning as speech technology evolves
- Work-life balance (on-call is light and well-supported)

## Interview Process

Our interview process typically takes 2 weeks.

- **Step 1: Recruiter Screen** (30 min)
  Background discussion and role overview.

- **Step 2: Technical Screen** (60 min)
  Speech fundamentals with a speech engineer.

- **Step 3: Speech Deep Dive** (90 min)
  Technical discussion about ASR architectures, your experience, and approach to speech challenges.

- **Step 4: ML Coding** (90 min)
  Practical exercise—audio processing, model implementation, or evaluation.

- **Step 5: System Design** (60 min)
  Design a speech processing system: streaming ASR, diarization, or voice pipeline.

- **Step 6: Team Interviews** (2 x 30 min)
  Meet potential teammates. Discussion about collaboration.

---

*[Company] is an equal opportunity employer.*