AI LLM / NLP Tech Lead (On-site) — Islamabad
Get AI-powered advice on this job and more exclusive features.
Start : ASAP
About VulSight
Location : Islamabad, Pakistan — on-site, full-time
Team : Engineering (AI / Research)
Start : ASAP
VulSight is a Web3 security firm helping teams ship safely across Move, Rust, and EVM ecosystems. We’ve delivered hundreds of assessments in traditional cybersecurity and Web3, and we’re now building NLP / LLM-driven agents that reason over code, audits, and on-chain data to catch bugs earlier and accelerate reviews.
The Role
You’ll own our AI roadmap end-to-end— research → prototyping → productization → team leadership . Your focus is on LLMs, NLP, and AI agents (tool use, planning, retrieval, evaluation) grounded in solid ML fundamentals . Blockchain / security knowledge is a plus.
What you’ll do
- Design & ship agentic systems that reason over smart-contract code (Move / Solidity), diffs, on-chain data, and prior audits; perform retrieval, analysis, test generation, and triage.
- Build RAG pipelines (indexing, chunking, retrievers), long-context strategies, tool calling, and multi-step planning (e.g., LangGraph-style flows).
- Own LLM evaluation : golden sets, style metrics, rubric-based grading, and production feedback loops (hallucination detection, safety, guardrails).
- Lead prompt engineering and model customization (SFT, LoRA / QLoRA, instruction tuning, RLAIF / RLHF where appropriate).
- Stand up inference infrastructure (vLLM / TGI / Triton), latency & cost optimization, caching, and observability (tracing, telemetry, error budgets).
- Curate high-quality datasets from public repos, audits, and on-chain sources; build labeling and synthetic-data pipelines.
- Collaborate with security researchers to turn workflows into auditor-grade copilots (invariant mining, property / test generation, patch suggestions).
- Hire, mentor, and lead a small on-site AI team; drive roadmap, rituals, and delivery quality.
Must-have qualifications
Strong Python and at least one major DL framework (PyTorch preferred).4–8+ years in ML / AI; 2+ years owning end-to-end ML / LLM projects or leading small teams.Hands-on with LLM application stacks : retrieval (FAISS / Milvus / pgvector), embeddings, tool use / function-calling, streaming, telemetry.Experience with agent orchestration (e.g., LangGraph, LangChain, LlamaIndex—or custom graphs) and productionizing LLM apps.Solid NLP fundamentals (tokenization, sequence modeling, evaluation), plus prompting & instruction-tuning best practices.MLOps / inference : serving (vLLM / TGI), batching, quantization, containerization, CI / CD for models, and experiment tracking.Clear product sense : turning fuzzy research ideas into shippable features with measurable impact.Nice to have
Blockchain / Web3 familiarity (Move, Solidity, Rust), static / dynamic analysis, or prior security / audit exposure.Program analysis / compilers / AST work; code-intelligence or LLM-for-code experience.Fine-tuning at scale (LoRA, PEFT), dataset distillation, or RLHF / RLAIF pipelines.Systems performance chops (CUDA / Triton, vectorization) or search / retrieval research.Experience with on-chain data tooling (Aptos / Sui / EVM nodes, indexers, subgraphs).Why join
Real users, real problems : your work lands in auditors’ hands immediately.Ownership : greenfield architecture with the mandate to do it right.Learning loop : tight feedback with senior security researchers and real codebases.Compensation
Competitive salary + equity . Hardware and learning budget provided.
Seniority level
Mid-Senior levelEmployment type
Full-timeJob function
Engineering and Information TechnologyIndustriesBlockchain Services#J-18808-Ljbffr