AI Engineering Lead

TRANZACT is a leading direct-to-consumer insurance services business, specializing in the distribution of Medicare Advantage, Medicare Supplement and Life & Supplemental insurance policies.  Our focus is on leveraging our highly skilled team of over 3,000 professionals to deliver innovative solutions for our insurance carrier partners and provide consumers with a best-in-class experience. 

We work with some of America's largest insurance brands to attract, educate and connect with consumers.  Our diverse team of analysts, writers, engineers, designers, business leads, data scientists and sales professionals all work together to create thoughtful, data-driven and effective marketing, technology, sales and servicing strategies.  These efforts have resulted in TRANZACT earning recognition and winning awards from some of the biggest names in the industry.

When you become part of TRANZACT, you join a team of passionate people who push for excellence, follow through on commitments, and love to succeed together.  Our culture rests on a simple principle – "Be Real" – that means being genuine, keeping your word, and having the courage to make tough decisions.  Our "Be Real" values resonate and are meaningful to us.  We are entrepreneurial, fast paced and empower our people.  We know how to make things happen and know that our people are always the source of our success.

About the Role

We are hiring a hands‑on AI Engineering leader to design and scale TRANZACT's next‑generation Conversational & Composable AI stack across three pillars: Voice AI, Agentic AI, and AI‑driven Analytics. You will own systems‑level architecture and delivery—from low‑latency speech and telephony to agentic orchestration and governed analytics—turning unique data assets into durable business outcomes. This role combines strategic ownership with sleeves‑rolled‑up execution.

What You'll Own (Systems‑Level Outcomes)

• Define and evolve the end‑to‑end Conversational AI platform: realtime ASR → agentic reasoning → TTS with sub‑second turn‑taking and production‑grade reliability.
• Establish scalable multi‑agent patterns (task routing, tool use, memory, human‑in‑the‑loop) and standard SDKs/templates to accelerate team adoption.
• Stand up enterprise‑grade Conversational RAG: hybrid retrieval, grounding/citation, freshness, and streaming context for voice/chat.
• Implement Responsible AI controls, observability, and governance (catalogs, lineage, model registry, policy enforcement) across the stack.
• Uplift the broader Engineering org by sharing reusable components, golden paths, and best practices; reduce vendor dependency via self‑service capabilities.
• Serve as Lead Engineer/Producer for at least one critical initiative per quarter, driving requirements → architecture → implementation → launch → post‑launch learning.

Key Responsibilities

• Architect low‑latency streaming pipelines and telephony/queue integrations (SIP/IVR/contact center) to support overflow, warm transfer, and compliant prompts.
• Design reference architectures and SDKs for Voice AI, Agent Assist, QA & Monitoring, and Conversational Analytics; drive adoption via docs and enablement.
• Build fine‑tuning pipelines (data curation, SFT, preference/reward optimization, safety tuning) and automated eval gates prior to promotion.
• Define eval suites (faithfulness, answer relevance, context recall, ASR WER, latency budgets) and instrument tracing/attribution/redaction across the retrieval/generation chain.
• Partner with Data/Platform for governed deployment: role‑based access, audit trails, change capture, freshness SLAs, and model/asset lifecycle management.
• Lead vendor benchmarking and make build/buy/insource recommendations aligned to unit economics, reliability, and IP strategy.
• Mentor engineers/data scientists; review designs, prompts, datasets, and PRs; host tech talks and share reusable components across teams.

Success Profile (6–12 Months)

• Conversational AI platform reliably supports key production use cases with clear SLOs and runbooks.
• Reusable SDKs/templates and "golden paths" enable faster delivery across teams; internal adoption measurably increases.
• Enterprise Conversational RAG delivers grounded answers within voice turn‑taking budgets, with healthy eval & monitoring signals.
• Responsible AI controls and governance are standardized and auditable across Voice/Agentic/Analytics workloads.
• A critical initiative is led to launch with documented impact and a retrospective feeding the platform roadmap.

Minimum Qualifications

• 7+ years in applied ML/AI or realtime distributed systems; 3+ years leading production LLM/voice solutions.
• Proven experience building realtime voice agents (WebRTC‑class streaming or equivalent) with measurable business impact.
• Hands‑on fine‑tuning experience (parameter‑efficient and, when needed, full‑parameter): data curation, SFT, preference/reward optimization, safety tuning, and evals.
• Production RAG for conversational use cases (hybrid retrieval, reranking, caching, grounding/citation) with strong observability.
• Strong engineering in Python and PyTorch; exposure to TypeScript or similar is a plus.
• Bachelor's degree in Computer Science required; Master's in CS (AI focus) preferred.
• Experience with governed data/ML platforms (catalog, registry, lineage) and secure deployment patterns.
• Track record of mentoring and uplifting teams; effective communicator with stakeholders across technical and non‑technical domains.

Preferred Qualifications

• Vector/search platforms and rerankers; hybrid lexical + dense retrieval; multilingual/domain adapters.
• LLM traces/observability tooling; model registries; dataset versioning and approval gates.
• Safety tooling (policy classifiers, redaction/PII handling) and progressive rollout (shadow/canary/feature flags).
• Streaming RAG for Voice AI (incremental retrieval, early‑token generation/TTFB optimization, barge‑in‑friendly generation).
• Telephony/contact‑center systems and marketing analytics experience.

How We Work

• Composable AI first: best‑in‑class components while protecting proprietary data/process IP.
• Agile and iterative: prove value quickly, then scale with governance and reliability.
• Responsible by default: industry frameworks, content controls, lineage, permissions, and continuous evaluation.

At TRANZACT we promote inclusion and job insertion without distinction based on disability, gender, orientation, gender identity and expression, place of origin or any other reason protected by national legislation. The company respects diversity and does not tolerate acts of discrimination in its processes.

Como empresa, creemos en la igualidad de oportunidades e inclusión. Consecuentemente, esta vacante de trabajo está disponible para personas con discapacidad (de acuerdo a lo indicado en la Ley No