Arjun Mehta
arjun.mehta@example.com+44 20 7946 0202linkedin.com/in/arjunmehtagithub.com/arjunmehta
Summary
AI engineer with 6 years building large language model applications, retrieval pipelines, and production inference systems.
Skilled in PyTorch, RAG architecture, prompt engineering, vector databases, evaluation harnesses, and shipping generative features that hold up under real traffic.
Experience
Senior AI Engineer
Cognita Labs | London, UK | March 2022 - Present
- Lead a 4-engineer team building a retrieval-augmented assistant answering 240,000 customer queries per month.
- Designed a hybrid retrieval pipeline over a pgvector store that improved answer relevance by 31% in offline evaluation.
- Built an LLM evaluation harness with grounded-answer and hallucination checks, raising factual accuracy from 78% to 91%.
- Optimised inference with batching and quantisation, cutting median response latency from 4.2 s to 1.6 s.
- Mentor 3 engineers on prompt design, evaluation methodology, and safe rollout of generative features.
AI Engineer
Verbal Health | Cambridge, UK | June 2019 - February 2022
- Built clinical-note summarisation models fine-tuned in PyTorch on 180,000 de-identified records.
- Implemented a guardrail layer for prompt injection and PII redaction that blocked 99.4% of flagged inputs.
- Developed a vector search service over FAISS serving 1,200 queries per second for symptom triage.
- Reduced annotation cost by 38% through active learning and weak supervision on labelling pipelines.
Machine Learning Engineer
Lexio | Manchester, UK | September 2017 - May 2019
- Trained text classification models in Python and scikit-learn for document routing across 9 categories.
- Built feature pipelines and embedding caches that reduced model serving cost by 27%.
- Shipped an A/B tested re-ranking model that improved search click-through by 14%.
- Automated model retraining in Airflow, removing 8 hours of manual release work per week.
Skills
Languages
Python, TypeScript, SQL, Bash
AI
LLMs, RAG, Prompt engineering, PyTorch, Fine-tuning
Retrieval
pgvector, FAISS, Pinecone, Embeddings, Hybrid search
Platform
AWS, Docker, Kubernetes, LangChain, Weights & Biases
Education
- MSc Data Science, University of Edinburgh
- BSc Mathematics, University of Leeds