Mar 2023 – Mar 2024 · 1 yr

• Team lead for the AI team at Staple AI. My work primarily focuses on:
• 1. Building Natural Language Processing and Computer Vision Solutions for Document Processing using Neural Networks (Specifically LLMs & CNNs).
• 2. System Design focused on deploying ML solutions on multiple cloud environments (AWS, GCP, Azure).
• 3. ML-Ops supervision for efficient LLM deployment on cloud services and on-premise.
• 4. Managing a team of 6 talented full time engineers to drive projects across NLP-Research, Backend Development (Python / Rust).

Jun 2021 – Mar 2023 · 1 yr 9 mos

• Building the next generation of AI-Driven Document Processing products.

Feb 2021 – May 2021 · 3 mos · Bengaluru, Karnataka, India

• 1. Ideated and built a Proof of Concept for an MRR/ARR based financing for subscription based companies using DeFi banks and protocols (Centrifuge and Maker DAO). Conducted user research interviews with over 45 SAAS founders.
• 2. Developed an algorithm for automated at home screening of primary care health problems using Natural Language Processing and Image Processing Algorithms. Partnered with doctors from over 4 hospitals to research potential use cases.
• 3. Developed user personas, a product roadmap, market sizing stats, various business models and a technical scope document.
• EF is the world’s leading talent investor. 3% acceptance rate. Backed by Founders Fund and Greylock Partners with Reid Hoffman (Founder, LinkedIn) and Demis Hassabis (Founder, Deep Mind) on the board.

May 2020 – Feb 2021 · 9 mos

• 1. Revamped the User-Experience for EY-Cognistreamer to reduce churn by over 45% through various parts of the product.
• 2. Designed and implemented a low-code Machine Learning Model (NLP & CV) deployment platform. The platform reduced deployment time by 85%. [CI/CD automation, Docker, AquaSec, Kubernetes, Tensorflow, PyTorch].
• 3. Implemented a distributed microservice logger to monitor health, execution metadata, and metrics of deployed services and display on a PowerBI dashboard. [Stream Analytics, Azure App Insights]
• 4. Implemented a testing framework to automatically test new APIs created via the low-code tool. [Selenium, Requests]

Feb 2019 – Apr 2020 · 1 yr 2 mos

• 1. Architected an active learning solution responsible for collecting user data, implementing data sanity checks, retraining the model and redeploying the new model. [SQL, Pandas, Django, Docker, Kubernetes, Azure Container Registry]
• 2. Built a solution for inferring cross-lingual semantic similarity in text documents with a custom distance function. [Keras]
• 3. Designed a content-based recommendation system to connect employees with project opportunities. [Tensorflow]

May 2018 – Aug 2018 · 3 mos

• As a developer for BRL-CAD, I :
• 1. Constructed python bindings for interacting with BRL-CAD’s source code written in C.
• 2. Established a standard library of abstractions around the bindings.
• 3. Designed and implemented BRL-CAD's python procedural geometry engine to automate complex geometry generation.

Apr 2018 – Jan 2019 · 9 mos · Hyderabad Area, India

• Worked with one of the largest medical firms in the world to
• 1. Design and develop solutions to automate tracking of Adverse Drug Effects in medical text files.
• 2. Save over 3000 man hours per week that goes into manual analysis of documents.
• 3. Develop a tool for machine assisted reading to ensure that no information is missed out when dealing with information rich documents.

Apr 2016 – Aug 2016 · 4 mos

• Worked with Apertium for Google Summer of Code to develop a lint program which tests Apertium data files for suspicious or unrecommended constructs (which most likely grow into bugs)

Jun 2015 – Aug 2015 · 2 mos · Gurgaon, India

• Worked as a communications intern at one of France's most successful start-up . A part of the India team organizing the Bla Bla Tour - 2015 , India. As an intern at BlaBlaCar, my main mantle revolved around working on social media outreach, organizing local events pan India and working with the community management team for handling active client issues