Opkey
Website:
opkey.com
Job details:
Opkey is inviting applications for the position of VP of AI.
Job Role—VP of AI—Opkey
Location: Bangalore, India (Hybrid)
Function: AI & Engineering
About Opkey
Opkey is the leading Cloud Application Lifecycle Management (CALM) platform for Oracle, Workday, Salesforce, Coupa, and more. It cuts the costs and risks that drag down implementations and ongoing change, helping you go live on time, get more from your cloud app investments, and reach AI readiness faster. Opkey's 20+ AI agents manage all five phases of the cloud application lifecycle—Define, Design, Configure, Test, and Train.
Whether it's a new implementation, a platform update, or a business-as-usual change, Opkey handles it all: updates validated in hours, self-healing tests, end-to-end integrations assured, configurations synced, and training updated in real time—all delivered in a single unified platform instead of a patchwork of disconnected tools.
Powered by Argus, a domain-specific AI model trained on decades of expertise and terabytes of enterprise application data, Opkey automates configuration, testing, change impact analysis, and training across these applications—cutting manual effort by 80%, enabling 30% faster go-lives, and slashing downtime risk by 92%.
The Opportunity
We are looking for an exceptional AI Leader to own and evolve the intelligence layer that powers Opkey's entire platform. This is a rare opportunity to lead AI at a company where AI isn't a feature—it's the product.
You will own two deeply interconnected missions: the continuous improvement of Argus, our proprietary domain-specific Small Language Model (SLM), and the architecture and evolution of the agentic framework that orchestrates 20+ AI agents across the cloud application lifecycle. You will work at the intersection of applied research, engineering, and product—and your work will directly determine how fast Opkey can automate, how accurately it can reason over enterprise data, and how far ahead of competitors it stays.
This is a hands-on leadership role. We want someone who can think at the level of research and strategy and also get into training pipelines, agent architectures, and evaluation frameworks when the work demands it.
What You Will Own
Argus—Domain-Specific SLM
- Drive the continuous pretraining, fine-tuning, and post-training (RLHF/DPO/SFT) of Argus on enterprise application data—Oracle, Workday, Salesforce, Coupa, and more
- Build and own the data flywheel: curating, labeling, and synthesizing high-quality domain-specific training data from Opkey's growing corpus of enterprise application signals
- Design and implement evaluation frameworks to measure model quality, reliability, and hallucination rates on domain-specific benchmarks
- Own model efficiency—quantization, distillation, and inference optimization to ensure Argus runs fast and cost-effectively in production
- Build a culture of continuous improvement: establish processes to systematically identify where Argus is underperforming and close those gaps Agentic Framework
- Architect and evolve the multi-agent orchestration layer that powers Opkey's 20+ AI agents across Define, Design, Configure, Test, and Train phases
- Define how agents plan, reason, use tools, and hand off to each other in complex multi-step enterprise workflows
- Build reliability and observability into the agent layer—guardrails, fallback logic, tracing, and evaluation in production
- Drive the integration of Argus into agent workflows, determining where the SLM handles tasks vs. where it defers to larger foundation models
- Stay ahead of the field: evaluate and adopt emerging agentic patterns (MCP, multiagent coordination, long-context reasoning) that can give Opkey a competitive edge.
Leadership & Team Building
- Build and lead a high-performing team of Applied Scientists, ML Engineers, and AI researchers
- Define the AI research and engineering roadmap in partnership with Product and CTO
- Represent Opkey's AI capabilities externally—with customers, partners, and in the broader AI community
- Establish the hiring bar and culture for AI at Opkey
Must-Have Experience
What We Are Looking For
- 10+ years in AI/ML, with at least 3–4 years in a leadership or principal-level role
- Hands-on experience training or fine-tuning language models (LLMs or SLMs) — not just consuming APIs, but owning the training pipeline end-to-end
- Deep familiarity with post-training techniques: RLHF, DPO, SFT, instruction tuning
- Experience building or leading agentic AI systems—multi-agent orchestration, tool se, planning and reasoning frameworks
- Strong command of the modern ML stack: PyTorch, Hugging Face, PEFT/LoRA, vLLM or equivalent inference frameworks
- Experience with evaluation design for LLM/SLM systems—building domain-specific benchmarks, red-teaming, hallucination measurement Highly Valued
- Experience at companies doing serious model training work in India
- Background in enterprise software, ERP, or cloud applications (Oracle, Workday, Salesforce)—or experience building AI for vertical/domain-specific use cases
- Experience with synthetic data generation and data curation at scale
- Publications or patents in NLP, LLMs, or agentic AI (a plus, not a requirement)
Leadership Qualities
- You can zoom out and set a 12-month AI roadmap, and zoom in and debug a training run.
- You hire for depth and build teams that raise each other's bar.
- You communicate complex AI tradeoffs clearly to non-technical stakeholders.
- You have strong opinions, loosely held—and you update fast when the evidence changes.
Why This Role?
- Argus is real. This isn't a role where you're building an AI wrapper on top of GPT. You're improving a proprietary domain-specific model trained on unique enterprise data that no one else has.
- The agentic system is live. 20+ agents are already in production. You're evolving a system that's working, not starting from scratch.
- The domain is defensible. Enterprise cloud application data—test cases, configurations, change logs, implementation patterns—is not something OpenAI or any foundation model is trained on. This is Opkey's moat, and you'll own it.
- Your work ships. Opkey serves real enterprise customers. Improvements you make to Argus and the agent framework translate directly into customer outcomes—faster go-lives, fewer failures, and less manual work.
- The timing is right. Agentic AI for enterprise is still early. The decisions made in the next 12–18 months will define Opkey's AI architecture for years. You'll make those decisions.
Skills: dpo,sft,agentic ai,llm,rlhf,pytorch,slm,vllm
Click on Apply to know more.