About Repello AI

Repello AI is an AI Security company helping businesses take GenAI products from PoC to production safely and securely. As organizations embed LLMs into critical workflows, new attack surfaces emerge—prompt injection, jailbreaks, data leakage & exfiltration, model abuse, and compliance gaps. Financial, regulatory, and reputational stakes are high; we help companies identify vulnerabilities and secure their AI offerings before attackers do.

We're early, well-backed, and building a product-first culture where security, reliability, and speed coexist.

Why This Role Matters

Today, many enterprise GenAI teams "just prompt the model" and ship. That breaks at scale: latency spikes, context windows overflow, costs balloon, and security fragility leaks data. We need an AI Engineer with strong software engineering + distributed systems instincts who can turn prototype LLM workflows into fast, reliable, observable, and hardened production pipelines—and do it as part of our founding team.

You will own the core LLM / RAG / agent infrastructure that powers Repello AI's security platform and helps customers harden theirs. If you like solving real performance bottlenecks, building retrieval systems that actually work in prod, and instrumenting guardrails that stop prompt attacks, you'll love this role.

What You'll Do

Build & optimize GenAI/LLM pipelines that move beyond brittle system prompts to structured, testable, monitored workflows.

End-to-end ownership: ingestion → chunking/embedding → vector/hybrid retrieval → context assembly → model routing/inference → post-processing → eval/guardrails → observability.

Prompt & template management: parameterized prompts, versioning, A/B and automated evals; enforce structured / JSON outputs.

Latency + cost optimization: caching (embedding & generation), adaptive context pruning, async batch calls, streaming, dynamic model selection, distillation/fine-tune vs API trade-offs.

Security & robustness hardening: detect/mitigate prompt injection & jailbreak attempts; input/output filtering; secrets isolation; red-team corpora replay; adversarial content fuzzing.

Evaluation & observability: build metrics for retrieval quality, attack success rate, hallucination, token/latency/cost dashboards; integrate traces (OpenTelemetry, LangSmith, etc.).

Agent & tool orchestration: integrate tool-calling, function-calling, and Model Context Protocol (MCP) style capabilities to let models safely call scanners, scanners call models, and customers plug in custom tools.

Customer-facing engineering: work with design partners to productionize secure GenAI patterns inside their environments; contribute feedback loops that shape our product roadmap.

Founding responsibilities: help define tech strategy, influence hiring, shape engineering culture, and set quality/security bars.

Minimum Qualifications

• Solid software engineering background (data / backend / platform). Typically 3+ years production experience; strong candidates with ~2 years plus serious OSS/startup depth welcome.

• Hands-on with LLM or GenAI features in production: shipping user-facing flows, automations, or internal tooling that call foundation model APIs or self-hosted models.

• Vector database + retrieval experience with at least one of: Pinecone, Weaviate, Qdrant, Milvus, pgvector/PGVector, OpenSearch k-NN, FAISS, Chroma.

• Python proficiency plus experience in one additional production language (TypeScript/Node, Go, Rust, or similar).

• Cloud + containers: AWS/GCP/Azure; Docker; Kubernetes or managed container services; infra-as-code basics.

• Data engineering fundamentals: ETL/ELT, streaming/batch ingestion, schema & metadata management.

• Familiar with one or more LLM / agent / orchestration frameworks (LangChain, LlamaIndex, LangGraph, CrewAI, Haystack, Semantic Kernel, custom in-house, etc.).

• Strong debugging skills across distributed systems; comfortable instrumenting traces & metrics.

• Excellent communication; able to translate product/security needs into technical build plans.

Preferred / Nice-to-Have

• Experience building LLM guardrails: jailbreak filtering, prompt injection detection, output sanitation, PII redaction.

• Prior work in AI red-teaming, adversarial ML, or application security.

• Familiarity with model eval tooling (DeepEval, OpenAI Evals style harnesses, custom test corpora) and automatic regression testing.

• Experience with structured outputs & schema enforcement (JSON mode, function calling, tool invocation contracts).

• GPU orchestration / model serving (vLLM, TensorRT-LLM, Ray Serve, KServe, Sagemaker, Vertex AI, etc.).

• Contributions to OSS in GenAI / security / tooling; conference talks, blogs, or papers.

Why Join Us

• High impact, early seat on the founding engineering team; shape the core platform & culture.

• Build in a space that matters: securing the GenAI layer before attackers do.

• Fast-paced, learning-heavy environment with direct customer exposure.

• Competitive salary + meaningful equity.

• Backed by strong investors; building for global adoption from Day 1.

How to Apply

Email naman@repello.ai with:

1. Resume / CV.

2. Short note: Why Repello AI? Why this role? (a few paragraphs is fine.)

3. Links to relevant work: GitHub, demo repo, blog post, package, talk, or shipped feature.

4. (Optional) Quick Loom / video walkthrough (≤5 min) of something you built.

If you're unsure whether you fit every bullet—apply. We're looking for builders who care about reliability, security, and shipping.

A Note on Diversity & Inclusion

We value a diversity of backgrounds, perspectives, and experiences. If you’re excited by our mission but don’t meet every qualification, we still encourage you to reach out.