Applied Scientist, GenAI

TraceLink•Wilmington, MA

3d•Onsite

About The Position

TraceLink’s software solutions and Opus Platform help the pharmaceutical industry digitize their supply chain and enable greater compliance, visibility, and decision making. It reduces disruption to the supply of medicines to patients who need them, anywhere in the world. Founded in 2009 with the simple mission of protecting patients, today Tracelink has 8 offices, over 800 employees and more than 1300 customers in over 60 countries around the world. Our expanding product suite continues to protect patients and now also enhances multi-enterprise collaboration through innovative new applications such as MINT. Tracelink is recognized as an industry leader by Gartner and IDC, and for having a great company culture by Comparably. Staff / Senior Applied Scientist, GenAI & ML Systems Location: Wilmington, MA (US) - Fulltime Onsite About the Role We are hiring a Staff / Senior Applied Scientist to lead the design and deployment of production-grade GenAI and ML systems with a strong emphasis on being hands-on. You will personally build, iterate, and ship systems focused on LLM/SLM optimization for agentic, multi-agent architectures in cloud environments. This role is ideal for someone with deep expertise in one or more areas of LLM/SLM optimization for agent-based systems, and hands-on experience in designing, implementing, and operating large-scale multi-agent systems in the cloud.

Requirements

Context engineering for complex multi-agent systems (prompt orchestration, tool calling, memory/state design, routing, constraint handling)
Fine-tuning of SLMs and delivering them to production (training strategies, validation, deployment, monitoring, rollback readiness)
Experience with Advanced RAG, semantic search, embeddings, and cross-encoders (retrieval tuning, chunking strategies, query rewriting/planning, reranking)
Ability to translate ambiguous requirements into concrete architectures, metrics, and deliverables
Hands-on inference optimization experience: quantization, distillation, batching, caching, model routing, speculative decoding
Experience building retrieval systems at scale using vector DBs and search stacks
Comfort working across the full lifecycle: research → prototype → A/B test → production hardening

Nice To Haves

Familiarity with enterprise constraints: privacy, security, data governance, permissions, auditability
Experience designing and running GenAI observability: traces, prompt/versioning, tool call logging, feedback loops
Strong ability to implement production-quality systems in Python (and/or adjacent backend languages)
Proven experience deploying GenAI/ML systems in cloud environments (AWS/Azure/GCP)
Experience with scalable inference and service operations: containers, APIs, observability, reliability practices
MS/PhD in CS/ML/NLP/Stats (or equivalent applied experience building production systems)

Responsibilities

Hands-on ownership of building and shipping multi-agent systems (planner/executor, tool-using agents, supervisor patterns, routing, role-based agents) from prototype to production.
Write production-quality code for agent orchestration, tool integration, memory/state design, and context management.
Lead context engineering strategies for multi-agent coordination: prompt design, state persistence, agent handoffs, grounding, constraints, and safety controls.
Hands-on fine-tune and deploy SLM models for production usage: dataset creation, training workflows, evaluation, and inference serving.
Build Advanced RAG pipelines end-to-end, including semantic search, embeddings, hybrid retrieval, and cross-encoder reranking.
Implement evaluation frameworks for multi-agent systems covering quality, latency, cost, robustness, and failure mode detection.
Collaborate with platform and product engineering to ensure solutions are cloud-native, secure, observable, and scalable (monitoring, logging, CI/CD).
Optimize for cost and latency via model routing, caching, compression strategies, and inference efficiency improvements.
Mentor peers through code reviews, architecture sessions, and hands-on technical leadership.