Founding AI Engineer

HalluSeer Labs · Founding team · Equity

The company. AI voice agents have crossed into production at banks, healthcare providers, and BPO call centers — and no one independently audits what they say to customers. HalluSeer receives completed calls via webhook from seven voice platforms (Retell, Vapi, NLPearl, ElevenLabs, Bland, Synthflow, Sierra), detects hallucinations and policy violations against each client's own knowledge base, and routes flagged calls to a network of expert human validators. Providers can't credibly grade their own output; regulated buyers increasingly can't deploy without someone who does.

What you'll own — the hole in the middle. The detection engine. The schema is deployed and waiting (event taxonomy, scoring, HITL queue, validator marketplace — all built, all empty). The work is sequenced in a 17-task technical plan you'd be free to challenge:

• Transcript normalization across seven provider formats
• Speaker attribution (you can only judge what the agent said)
• Retrieval-grounded hallucination detection (LettuceDetect-class span models) against the client's own corpus
• Independent ASR verification — re-transcribe provider audio and diff it, catching errors providers structurally cannot see
• PII detection feeding both compliance events and redaction
• Deterministic risk scoring, a customer-authorable rules engine, HITL routing
• The feedback loop: paid expert reviews become labeled training data — the long-term moat

Plus: bringing engineering in-house from the current outsourced arrangement, and owning detection accuracy as a number you personally defend to customers.

You.

• Shipped ML or LLM systems to production and supported them afterward
• Know why precision beats recall when humans review model output
• Comfortable that the unglamorous first quarter — normalization, evaluation harnesses, golden sets — is what makes the model work real
• Strong plus: ASR/speech, RAG evaluation, or a compliance-adjacent domain (fintech, health, trust & safety)
• Strong plus: you've been the senior engineer somewhere that grew past the stage you enjoy

The CTO track — explicit, not implied. You join as Head of AI, the senior-most technical person in the company. The CTO title follows defined milestones, in writing in your offer: detection engine live with a design partner, measured accuracy holding an agreed floor, and the in-house team you've begun to build. No parallel CTO search runs while you deliver. If you hit the milestones, the title and expanded equity are yours; if we both discover it's not the right fit, you're still the founding engineer of the platform with founding equity.

The offer. Founding-level equity (standard 4-year vest, 1-year cliff), modest salary at close of the pre-seed currently in motion (anchor investor engaged), full remote, meaningful voice in company direction from day one.

Pay: $101,226.45 - $121,907.12 per year

Benefits:

• Flexible schedule

Education:

• Bachelor's (Preferred)

Experience:

• AI Engineering: 4 years (Required)
• AI Voice / Agentic: 2 years (Preferred)

Work Location: Remote