Gen AI Architect

Location:
Hybrid role in Warren, NJ

Duration:
Long Term Contract

Kindly share your resumes to dkannoji

About the Role: We are looking for a Senior GenAI / Small Language Model (SLM) Engineer / Architect to design, deploy, and maintain agentic AI solutions that are safe, scalable, and business ready. You will own end to end delivery—from prompt and agent design to data pipelines, model deployment, observability, and rigorous validation—partnering with product, architecture, security, and QA to ship AI features that perform reliably in production.hat

You’ll Do

SLM Design and Fine Tuning:

• Collect, clean, and preprocess domain-specific datasets for SLM training and fine-tuning. Ensure data quality, diversity, and compliance with privacy and security standards.
• Fine-tune small language models on curated datasets using techniques like LoRA, adapters, or parameter-efficient tuning. Optimize hyperparameters for performance, latency, and resource efficiency.
• Help design and implement agent orchestration (single and multi-agent) and function/tool use strategies. Craft, version, and optimize prompts and system instructions for accuracy, coherence, and domain alignment.
• Integrate external tools/APIs and establish content safety guardrails (e.g., policy enforcement, PII redaction, jailbreak prevention).

Implementation, Testing & Maintenance:

• Build resilient agent workflows and services; harden reliability with retries, fallbacks, circuit breakers. Develop automated tests for prompts, tools, and agent behaviors; maintain regression suites and golden datasets.
• Operate AI services in production: performance tuning, cost optimization, incident response, and iterative improvement.

Data & MLOps:

• Design and manage data pipelines for fine tuning and retrieval (RAG), including cleansing, labeling, and governance.
• Monitor drift, quality, latency, and safety signals; implement model/agent observability and alerting. Run structured evaluations of agent outputs (functional, coherence, safety, bias); track precision/recall and hallucination rates.
• Perform risk assessments for agent behaviors and tool actions; document mitigations and approval workflows. Collaborate with security/compliance to meet regulatory, privacy, and usage policy requirements.

Minimum Qualifications:

• 14+ Years of IT industry and 4–8+ years in software/ML engineering, with 2+ years building LLM/SLM/GenAI solutions in production. Proficiency in Python (and/or Type Script) and modern AI orchestration frameworks (e.g., Microsoft Agent Framework, Google Agent Development Kit, Lang Chain, Semantic Kernel). Hands on with retrieval augmented generation (RAG), function calling, prompt optimization, and agent design patterns.
• Experience building data pipelines (batch/stream), and managing datasets for training/fine tuning and evaluation. Practical understanding of AI guardrails: content filtering, safety policies, redaction, rate limiting, and misuse prevention. Strong willingness to learn advanced agent orchestration and MLOps practices.

Preferred Qualifications:

• MLOps fluency: model packaging, CI/CD, experiment tracking (e.g., MLflow), deployment on cloud/container platforms. IaC (e.g., Terraform/Bicep) and Dev Ops tooling (e.g., Git Hub Actions/Azure Dev Ops); strong grasp of observability. Experience with multi agent systems, toolformer patterns, and complex orchestration graphs.
• Knowledge of vector databases and retrieval systems; evaluation frameworks (e.g., Ragas, Deep Eval) and custom metrics. Familiarity with privacy, compliance, and model risk management practices for AI.
• Background in tuning open source and hosted models; comfort with hybrid cloud environments.

Tools & Technologies:

• Python;
  Type Script; MAF/Google ADK/Lang Chain/Semantic Kernel;
  Vector DBs and frameworks (e.g., Qdrant/FAISS/Pinecone); CI/CD (Git Hub Actions/Azure Dev Ops);
  IaC (Terraform/Bicep);
  Observability

Working Model:

• Partner with Product, Architecture, Security, and QA to plan, design, and ship safe AI features.
• Contribute to internal prompt standards, evaluation datasets, and reuseable components.
• Document designs, decisions, and risks; mentor peers and champion responsible AI practices.