EXPERIENCE

Past Roles

Jan 2023 - Present

SketchDeck.ai - Snr Machine Learning Engineer

Developed DAWN - ML service that detects structural steel elements, fetches their labels, and their lengths in a construction drawing with an accuracy of over 95% (tested across 750+ drawings), using state-of-the-art computer vision deep learning algorithms. This is helping over 140 estimators daily with their material takeoffs, saving them 65-70% of their time, generating over 1.2m in recurring yearly profits.
Responsible for designing, deploying, and maintaining the ML service DAWN on AWS, and GCP, ensuring that detection results delivered were under 3 minutes, and handling simultaneous loads from different customers without delays using an efficient messaging queue mechanism.
Used Generative AI prompt engineering to give customers a chatbot to clear doubts about the product, as well as answer any steel related questions. Used multimodal LLMs with few-shot prompting to understand various parts of pdf drawings which were harder for a general CV model. This saved the customers over 25-30% time and improved the overall workflow.

May 2022 - August 2022

Successfully migrated hardcoded broker limit overrides to the AWS AppConfig with proper change management and bar raiser strategies.
Developed a tool for the operators to update customers' broker limit overrides with just one command, enabling customers to get their increased limits in under a minute as opposed to the previous 10-day cycle.
Envisioned and coded a machine learning prototype tool for autoscaling of brokers based on the historical performance of customers' broker instance performances during the Hackathon week.

Sept 2020 - August 2021

Strategized the deployment and production of all flask-based ML applications without containerizing or hosting.
Extracted information for reconciliation in banks from random remarks and provided a recommendation API with structured information.
Set up an ensemble-based Image Processing pipeline for the extraction of details from various identification ids.
Created and deployed an automated testing framework to be used across multiple use-cases
Deployed an end-to-end Jenkins deployment job for automatically building, deploying, and mailing the results of the concerned ML project to all the stakeholders.
Built a UI supported application for customers to extract and verify signatures from any financial/personal documents using deep learning
Assisted my team in building a personal finance management tool that not only categorizes monthly expenditures but also provides predictions on what a particular user might spend next (Helpful for directed marketing)

July 2020 - Sept 2020

Built a text-based pokemon game called PokeWod on Discord using discord.py and psql
The game was deployed on GCP (currently deployed locally) and had more than 5000 active users until 2022 when I sold the code.
The overall complexity was kept down to O(n) and hence the response time of the bot is on an average 50MS.

July 2019 - July 2020

Predicted possible faults that may occur in both Heavy Duty and Light-Medium Duty vehicles for a Multinational Automobile Company.
Built a SAAS product - Chatbot which not only understands the users' typos but also their actual intention.
Led a team of 7 in revamping the replenishment model and providing business and predictive analytics of a retail company.
Built a SAAS product - Temperature Detector which can detect multiple people and analyze their temperature using the feed from a thermal camera.
Led a team of 5 in providing detailed analytics and predicted results for Mumbai's disasters and COVID cases on a dashboard.
Created a sentiment analyzer using the Bag of Words approach to calculate sentiment score for paragraphs written in Indian regional languages.

December 2016 - January 2017

Answered domain-related topics on their website. The answers were to be precise, yet interesting so that the users can relate and understand the basic concepts.
Received a certificate of Excellence for my writing skills.

May 2018 - July 2018

Interned in the IT department at NTPC Ltd. in SAP hardware and software
Coordinated in writing a Z-code (customized) from T-code (standard) on performance appraisals, part of HR-Module and secured a perfect score during viva and project report evaluation.

experience.py: Experience