Asks for

The ad says“Demonstrated leadership on technical tasks and/or technical teams, including Agile software development and leading one or more tasks to completion”
The ad says“Ability to design and run LLM evaluation, including task-specific benchmarks, golden datasets, LLM-as-judge methods, and tracing and observability tooling (LangSmith, Langfuse, Arize, Weights & Biases)”
The ad says“Strong proficiency in Python, including asynchronous programming and modern packaging and dependency management”
The ad says“Fluency with agentic frameworks and patterns such as LangGraph, LangChain, CrewAI, AutoGen, Pydantic AI, or vendor agent SDKs, including multi-step tool use, planning, memory, and state management, and with integrating models, tools, and data sources through open standards such as Model Context Protocol (MCP)”
The ad says“Fluency with agentic frameworks and patterns such as LangGraph, LangChain, CrewAI, AutoGen, Pydantic AI, or vendor agent SDKs, including multi-step tool use, planning, memory, and state management, and with integrating models, tools, and data sources through open standards such as Model Context Protocol (MCP)”
The ad says“Command of model adaptation techniques including fine-tuning, LoRA/PEFT, quantization, and distillation, and the judgment to know when adaptation is preferable to prompting or retrieval”
The ad says“Fluency with containerization and deployment frameworks such as Kubernetes or Docker, including GPU scheduling and resource management for training and inference workloads”
The ad says“Practical command of retrieval-augmented generation, including chunking and embedding strategy, vector databases (pgvector, Milvus, Qdrant, Weaviate, FAISS), and hybrid or re-ranked retrieval”
The ad says“Command of model adaptation techniques including fine-tuning, LoRA/PEFT, quantization, and distillation, and the judgment to know when adaptation is preferable to prompting or retrieval”
The ad says“Demonstrated ability to train and deploy machine learning models with PyTorch and the Hugging Face ecosystem (transformers, datasets, accelerate)”
The ad says“Fluency with containerization and deployment frameworks such as Kubernetes or Docker, including GPU scheduling and resource management for training and inference workloads”
The ad says“Fluency with agentic frameworks and patterns such as LangGraph, LangChain, CrewAI, AutoGen, Pydantic AI, or vendor agent SDKs, including multi-step tool use, planning, memory, and state management, and with integrating models, tools, and data sources through open standards such as Model Context Protocol (MCP)”
The ad says“Ability to design and run LLM evaluation, including task-specific benchmarks, golden datasets, LLM-as-judge methods, and tracing and observability tooling (LangSmith, Langfuse, Arize, Weights & Biases)”
The ad says“Fluency with agentic frameworks and patterns such as LangGraph, LangChain, CrewAI, AutoGen, Pydantic AI, or vendor agent SDKs, including multi-step tool use, planning, memory, and state management, and with integrating models, tools, and data sources through open standards such as Model Context Protocol (MCP)”
The ad says“Ability to design and run LLM evaluation, including task-specific benchmarks, golden datasets, LLM-as-judge methods, and tracing and observability tooling (LangSmith, Langfuse, Arize, Weights & Biases)”
The ad says“Comfort with Linux platforms and command-line environments”
The ad says“Command of model adaptation techniques including fine-tuning, LoRA/PEFT, quantization, and distillation, and the judgment to know when adaptation is preferable to prompting or retrieval”
The ad says“Practical command of retrieval-augmented generation, including chunking and embedding strategy, vector databases (pgvector, Milvus, Qdrant, Weaviate, FAISS), and hybrid or re-ranked retrieval”
The ad says“Proven ability to serve models in production with inference frameworks such as vLLM, TensorRT-LLM, Triton Inference Server, Ollama, or ONNX Runtime, including latency, throughput, and cost tradeoffs, as well as deployment to edge or resource-constrained environments, including on-device inference”
The ad says“Proven ability to serve models in production with inference frameworks such as vLLM, TensorRT-LLM, Triton Inference Server, Ollama, or ONNX Runtime, including latency, throughput, and cost tradeoffs, as well as deployment to edge or resource-constrained environments, including on-device inference”
The ad says“Strong proficiency in Python, including asynchronous programming and modern packaging and dependency management”
The ad says“Practical command of retrieval-augmented generation, including chunking and embedding strategy, vector databases (pgvector, Milvus, Qdrant, Weaviate, FAISS), and hybrid or re-ranked retrieval”
The ad says“A track record of building applications on top of large language models, including prompt engineering, structured output, and context management”
The ad says“Fluency with agentic frameworks and patterns such as LangGraph, LangChain, CrewAI, AutoGen, Pydantic AI, or vendor agent SDKs, including multi-step tool use, planning, memory, and state management, and with integrating models, tools, and data sources through open standards such as Model Context Protocol (MCP)”
The ad says“Proficiency with automated testing in Python (pytest), including regression suites for non-deterministic model and agent behavior”
The ad says“Strong proficiency in Python, including asynchronous programming and modern packaging and dependency management”
The ad says“Demonstrated ability to train and deploy machine learning models with PyTorch and the Hugging Face ecosystem (transformers, datasets, accelerate)”
The ad says“Practical command of retrieval-augmented generation, including chunking and embedding strategy, vector databases (pgvector, Milvus, Qdrant, Weaviate, FAISS), and hybrid or re-ranked retrieval”
The ad says“Command of model adaptation techniques including fine-tuning, LoRA/PEFT, quantization, and distillation, and the judgment to know when adaptation is preferable to prompting or retrieval”
The ad says“Working knowledge of server-side development (API definitions, REST services, streaming and asynchronous services, etc.)”
The ad says“Working knowledge of server-side development (API definitions, REST services, streaming and asynchronous services, etc.)”
The ad says“Proven ability to serve models in production with inference frameworks such as vLLM, TensorRT-LLM, Triton Inference Server, Ollama, or ONNX Runtime, including latency, throughput, and cost tradeoffs, as well as deployment to edge or resource-constrained environments, including on-device inference”
The ad says“Proven ability to serve models in production with inference frameworks such as vLLM, TensorRT-LLM, Triton Inference Server, Ollama, or ONNX Runtime, including latency, throughput, and cost tradeoffs, as well as deployment to edge or resource-constrained environments, including on-device inference”
The ad says“Practical command of retrieval-augmented generation, including chunking and embedding strategy, vector databases (pgvector, Milvus, Qdrant, Weaviate, FAISS), and hybrid or re-ranked retrieval”
The ad says“Proven ability to serve models in production with inference frameworks such as vLLM, TensorRT-LLM, Triton Inference Server, Ollama, or ONNX Runtime, including latency, throughput, and cost tradeoffs, as well as deployment to edge or resource-constrained environments, including on-device inference”
The ad says“Practical command of retrieval-augmented generation, including chunking and embedding strategy, vector databases (pgvector, Milvus, Qdrant, Weaviate, FAISS), and hybrid or re-ranked retrieval”
The ad says“Ability to design and run LLM evaluation, including task-specific benchmarks, golden datasets, LLM-as-judge methods, and tracing and observability tooling (LangSmith, Langfuse, Arize, Weights & Biases)”

AI/ML Engineer III

Technergetics — Utica/Rome, NY area

A Note About Our AI-Assisted Interview Process

We use an AI application, “Alex Taylor,” to conduct first-round interviews for this position. The interview takes approximately 35 minutes. If we would like to interview you, you will receive an email invitation from “Alex” within ten business days of your application.

Alex is available 24/7, which lets us conduct far more first-round interviews than our human staff's schedule alone would allow. Technergetics HR (and additional staff, as applicable) reviews every first-round interview. AI supports our decision-making, but all decisions about who advances to a first or second interview are made by our human staff. Candidates selected for a second-round interview will meet with the HR director and hiring manager.

Any data collected during the interview process, including AI-generated insights, is handled with care and confidentiality, in compliance with applicable data protection laws. Your responses are processed only to provide feedback on your skills and knowledge. Data is stored securely and will not be shared with third parties without your consent.

We understand some candidates may be hesitant to interview with an AI application — it is by no means perfect at this time. But as a company dedicated to research and development in AI/ML and other technologies, we see this as a chance to practice what we preach.

Opportunity Overview

Technergetics is looking for an AI/ML Engineer III to design, develop, and deploy advanced AI capabilities alongside a high-performing team of full-stack developers. This role centers on building production systems around foundation models, including agentic workflows, retrieval-augmented generation, and multimodal machine learning, for demanding government and commercial customers.

Contingent Position: This position is contingent upon contract award and funding.

Position Details

Salary Range: $125,000–$175,000 annually. The final offer depends on how many position qualifications the candidate meets, as well as education and experience. This is a full-time, exempt position.

Location, Travel, and Remote Work

Candidates who are located within, or relocate to, a commutable distance of the Utica/Rome area can expect to be onsite 20% of their workweek, for access to company and AFRL (Air Force Research Lab) facilities, secure data, and customers. A relocation signing bonus may be available.
Remote candidates outside a commutable distance to Utica/Rome will still be considered but may need to travel to the Utica/Rome area quarterly or more often, depending on company and client needs.
This position also involves approximately 5%–10% travel to customer and client sites outside the Utica-Rome, NY area.

Due to the security clearance required for this position, only U.S. citizens are eligible to apply, per Executive Order 12968 (Access to Classified Information).

Responsibilities and Duties

The successful candidate will work on one or more of our Machine Learning (ML) software products, with day-to-day activities that include:

Leading the design, development, and deployment of multi-modal machine learning architectures, including models and algorithms, to solve complex mission and business problems
Designing and building agentic systems on top of large language models for operational deployment
Implementing retrieval-augmented generation pipelines that ground model outputs against authoritative data sources
Defining and running evaluation for model and agent behavior in production
Optimizing model inference for production and edge/DDIL (denied, degraded, intermittent, and limited bandwidth) deployment scenarios
Applying AI assurance practices and documenting model limitations to support accreditation and customer review
Integrating state-of-the-art machine learning libraries, foundation models, and agent frameworks into existing software applications
Developing and maintaining data pipelines and supporting software for collecting, preprocessing, and transforming data for machine learning tasks
Designing software solutions, algorithms, and cloud architectures needed to satisfy product features and functionality defined by the product owner and other stakeholders in a production environment
Leading, coaching, and mentoring junior data scientists, engineers, and other staff
Contributing to phases of the software development life cycle, including functional analysis, technical requirements, technical design, prototyping, coding, testing, deployment, data migration, and support
Participating in daily scrums and working with the scrum master and scrum team to organize and prioritize workload through story-pointing, supporting delivery timelines and priorities
Collaborating with cross-functional teams to understand business requirements and translate them into machine learning solutions
Performing unit testing and debugging to identify and fix software defects, and contributing to code reviews with constructive feedback to peers
Staying current on new AI/ML approaches, frameworks, and industry trends
Serving as an AI subject matter expert for small teams of researchers and engineers on advanced R&D projects funded by government and/or commercial customers, and contributing to or leading proposal writing for new opportunities within your area of expertise

Education and Certifications

This position generally requires a Master’s degree from an accredited college or university in computer science, computer engineering, artificial intelligence, machine learning, or a closely related discipline. A Ph.D. in one of these fields is strongly preferred. A Bachelor’s degree in one of these fields, combined with seven or more years of directly relevant professional experience, will be considered in lieu of a Master’s degree.

Qualifications

Experience and Foundational Engineering

At minimum, three years of professional experience in machine learning or AI systems engineering, including at least one year building with large language models or other foundation models in a production setting
Strong proficiency in Python, including asynchronous programming and modern packaging and dependency management
Working knowledge of server-side development (API definitions, REST services, streaming and asynchronous services, etc.)
Fluency with containerization and deployment frameworks such as Kubernetes or Docker, including GPU scheduling and resource management for training and inference workloads
Hands-on work with at least one major cloud platform (AWS, Azure, or Google Cloud)
Comfort with Linux platforms and command-line environments
Familiarity with Continuous Delivery/Continuous Integration (DevSecOps, GitLab Pipelines, etc.)
Proficiency with automated testing in Python (pytest), including regression suites for non-deterministic model and agent behavior

Artificial Intelligence and Machine Learning

Demonstrated ability to train and deploy machine learning models with PyTorch and the Hugging Face ecosystem (transformers, datasets, accelerate)
A track record of building applications on top of large language models, including prompt engineering, structured output, and context management
Fluency with agentic frameworks and patterns such as LangGraph, LangChain, CrewAI, AutoGen, Pydantic AI, or vendor agent SDKs, including multi-step tool use, planning, memory, and state management, and with integrating models, tools, and data sources through open standards such as Model Context Protocol (MCP)
Practical command of retrieval-augmented generation, including chunking and embedding strategy, vector databases (pgvector, Milvus, Qdrant, Weaviate, FAISS), and hybrid or re-ranked retrieval
Ability to design and run LLM evaluation, including task-specific benchmarks, golden datasets, LLM-as-judge methods, and tracing and observability tooling (LangSmith, Langfuse, Arize, Weights & Biases)
Command of model adaptation techniques including fine-tuning, LoRA/PEFT, quantization, and distillation, and the judgment to know when adaptation is preferable to prompting or retrieval
Proven ability to serve models in production with inference frameworks such as vLLM, TensorRT-LLM, Triton Inference Server, Ollama, or ONNX Runtime, including latency, throughput, and cost tradeoffs, as well as deployment to edge or resource-constrained environments, including on-device inference
Grounding in AI safety and assurance practices, including guardrails, input and output filtering, prompt injection mitigation, and human-in-the-loop design
Direct work with multimodal models and cross-modal embedding across text, imagery, video, audio, or geospatial data

Leadership and Collaboration

Demonstrated leadership on technical tasks and/or technical teams, including Agile software development and leading one or more tasks to completion
Excellent communication and teamwork skills, including the ability to explain technical tradeoffs to non-technical stakeholders and customers

Nice to Have

Exposure to DoD cloud and software factory environments such as Platform One, BESPIN, AF Cloud One, DAF CLOUDworks, or AWS GovCloud, and with ATO and RMF processes at IL4/IL5
Familiarity with DoD and federal AI policy and governance, including CDAO Responsible AI guidance
Work with distributed training frameworks and schedulers (DeepSpeed, FSDP, Ray, Slurm)
Knowledge of knowledge graphs, ontologies, or Resource Description Framework (RDF), particularly as applied to graph-based retrieval (GraphRAG) and grounding
Background in developing modern full-stack web applications with frameworks such as Node, React, React Native, or Django
Basic working knowledge of Go, Java, C++, or other compiled languages
Contributions to open source AI/ML projects, or published applied AI research

Clearance

Selected applicants will undergo a security investigation and must meet and maintain eligibility for, at minimum, Top Secret access to classified information.

Benefits

Our benefits package includes health, life, disability, dental, and vision insurance, plus a 401(k) plan with a 3% company contribution and 3% company match.

Additional perks include:

Generous Paid Time Off, including a PTO “gift day” for your birthday
11 federal holidays per year
Three weeks of paid maternity/paternity leave
Annual technology allowance
Referral bonuses and professional recognition awards
Healthcare stipends
Tuition/education reimbursement (once eligibility requirements are met)
Flexible daily start and stop times for most projects and positions

Company Description

Technergetics is a U.S.-based company headquartered in Utica, NY, with employees and clients located throughout the country. The Utica/Rome area is a hub of cutting-edge cyber technology research, bolstered by the Griffiss Business & Technology Park's tenants and facilities, including the Air Force Research Lab (AFRL). At Technergetics, we work with a wide variety of technologies, including mobile and web apps, quantum computing, machine learning and artificial intelligence, AI-enabled edge devices, and more.

⚠ Beware of Fraudulent Job Offers and Postings

Technergetics will never extend an offer of employment without a thorough interview process that includes a face-to-face interview — either in person or via a virtual Teams meeting — from an official Technergetics email address (@techngs.com). If you receive correspondence from any other email address, it is a scam.

Technergetics does not discriminate on the basis of race, religion, color, national origin, gender, sexual orientation, age, marital status, veteran status, or disability status.