Website:
vikat.ai
Job details:
We’re looking for a AI Lead to own the AI architecture, strategy, and execution for our enterprise security product. You’ll lead a small team, design our Model Gateway, govern prompts, qualify models, and ensure AI makes us smarter without making us less trustworthy.
What You’ll Do
- Own the Model Gateway: provider registry, qualification suite, durable task queue, fallback strategies.
- Enforce prompt governance: thin prompts, thick code, pseudonymisation, broken chains, split audit.
- Build AI compute inside a hardware-isolated enclave: embedding encoder, prompt renderer, response validator.
- Drive business context extraction: closed questions into fixed schemas, promotion gates, confidence rules.
- Lead investigation and hunting: planner loop with typed tools, deterministic verdicts, narrative generation.
- Evaluate new architectures, track the model landscape, and recommend when to add or remove providers.
- Hire, mentor, and lead a team of AI engineers.
What You Bring
- 8+ years in software engineering, 3+ years leading AI/ML teams in production.
- Deep LLM integration experience (Bedrock, Azure OpenAI, Vertex, Anthropic, OpenAI).
- Strong understanding of model limitations: hallucination, prompt injection, data leakage, non-determinism.
- Production experience with embeddings and vector search (pgvector or similar).
- Experience building systems where AI is not on the critical path determinism, auditability, reproducibility.
- Security mindset: you understand why a security product can’t trust a model to make a decision.
- Enterprise deployment experience: customer-managed environments, sealed enclaves, compliance (SOC 2, ISO 27001).
- Excellent written communication for design docs and governance.
Nice to Have
- Confidential computing / secure enclaves (Nitro, Azure Confidential VMs, GCP Confidential Space).
- Security product background (SIEM, EDR, CSPM, attack path management).
- Graph databases (Apache AGE, Neo4j).
- MITRE ATT&CK familiarity.
- Open-source LLM inference (vLLM, TGI).
- Published work or talks on AI governance or AI in security.
Click on Apply to know more.