Join our Estimate Understanding team, part of our AI Review division. Large parts of the insurance industry run on scanned documents and PDFs; with some countries having 70% of all insurance claims made through scanned documents. Whenever an insurance company wants to repair a car; they receive an estimate from the body shop on how much it will cost to fix.The Estimate Understanding Team focuses on unlocking the rest of the insurance market for Tractable by building the estimate understanding layer. Our vision is to understand auto claims estimates in any format, in any language, as quickly as possible, so we can scale our product globally.One of the biggest challenges the team is working on is processing Japanese documents where there’s 10,000 characters in the alphabet and we’re required to be as accurate as a human transcribing the document and be able to scale to nation-wide volume.*Our key research problems right now include:** Detecting entities like body shop name in unstructured text in an image* Improving our text detection and recognition models to reach HLP on our documents* Improving NER on text that has been extracted by OCR*Our key engineering problems right now include:** Scaling to process tens of thousands of documents a month per a client* Creating better separation between our services* Reducing iteration time bringing new research into our serverless production service*As an Applied Researcher you will be responsible for:** Work on novel solutions to some of the biggest applied research challenges in Tractable* Work with the engineers on the team to deliver your improvements to production* Work on translating the teams long-term vision into a research roadmap* Presenting your research improvements and the real-world impact they have both internally and externally***What you need to be successful:** 3+ years of experience applying deep learning to complex real-world problem* 1+ year of Computer Vision specialisation* Passionate about bringing impact to the world through ML / CV* We train our models in PyTorch and deploy with torchscript* We’re a Python first team* Our production systems are primarily serverless through AWS Lambdas and StepFunctions* We often work with other services that use K8s and Kafka* Tractable runs everything on AWS***What’s in it for you:** Twice a year compensation reviews* Generous equity package* 25 days paid annual leave + bank holidays* 5% employer match on pension* Full private health coverage through Bupa* Access to a virtual GP through Babylon* Hybrid working (3 days working from home/week)* Generous maternity + paternity leave* Ability to work from abroad for up to 6 weeks/year* L&D budget to use on Learnerbly (our learning platform)* Mental health and career coaching through Sanctus* Headspace subscription* Payroll charitable donations (pre-tax)* Commuter Club* Cycle to Work scheme* £100 work from home budget-* Equity* Remote working/hybrid model: We‘re fully remote at the moment with no expectations of people to go into the office. We are moving to a hybrid model upon our return to the office in the Spring, where you‘ll only need to visit the office twice a week, our current office is based in Liverpool Street. Flexible working hours* Private Medical: (BUPA) - full coverage for yourself* Holidays: 25 days per year + bank holidays* Pension: 10% - 5% employee and 5% employer contributions* Tyve - As an employee you can take advantage of TYVE, a platform and service that allows you to donate part of their salary to a charity they want, directly from their payroll pre-income tax. The donation does come after National Insurance* Cycle to work scheme - Cycle Scheme, the UK’s most popular cycle to work benefit program. Tax is complicated. Cyclescheme isn‘t. It‘s a low effort, high impact employee benefit that gets staff cycling with savings of 25-39% on bikes and accessories. Employees pay nothing upfront, and the payments are taken tax efficiently from your salary by Tractable* Commuter Club - Commuter Club a scheme that provides cheaper Monthly/Weekly/Pay as you Go Travel, as well as big discounts of Annual tickets paying monthly via payroll. It caters to National Rail and City Transport systems including Oyster cards* Eye Test and Eye Wear Compensation through Expensify - Tractabler’s can take advantage of Tractable’s expense scheme for eyewear and eye tests. Employees can expense up to £100 on eye tests and eye wear from any branch or location* Headspace subscription* Classpass subscription* Sanctus Mental Health CoachingAccess* L&D Budget access through Learnerbly* 60 min HM interview with coding screen* Take home task* Final interview including ML & Stats, Take home task review and Values fit Computer Vision, Deep Learning, Python, Machine Learning, Artificial IntelligencePython, PyTorch, Tensorflow, Computer Vision, Deep Learning
This job has more than 30 days. You can find more up-to-date jobs using the search box.