Sumit Mandal

01/2023 – Current

Spearheaded and operationalized generative AI RAG bot via Open Source, generating daily savings of 0.39 Lakhs INR, totaling nearly 1.4 CR annually.
Engineered and deployed speech-to-text, translation, transcription, and summarization applications for customer support and sales teams, enhancing customer retention.
Developed RFQ and Pic to quote application using MultiModal Generative AI application, aiding in product upselling.
Led a team of 4 in creating a cost-effective ADAS solution, resulting in over 1 million Rupees savings for the company while ensuring life-saving capabilities.
Automated the conventional "Video – MER" authentication system, integrating Value Added Services like BMI Calculation and Disability Detection. Yielded a direct cost reduction of 24 CR.
Pioneered in live video and audio processing, leveraging untapped potentials in this domain.
Conceptualized and implemented CNN-based health insurance products, anticipating needs and preventing over 10,000 potential fatalities. Solely orchestrated end-to-end pipeline development.
Invited by Tata Motors to consult and enhance their ADAS solution.

06/2022 – 01/2023

Developed a fall detection system utilizing RNN, CNN, and Random Forest, deployed on Raspberry Pi integrated into smart helmets used by Adani and TATA, saving over 30 lives in hazardous conditions.
Engineered a highly optimized smart PPE Detection algorithm capable of handling 30+ simultaneous requests on a single server, leading to a significant 60% reduction in server usage.
Designed and implemented a Speech Detection model specialized in recognizing specific words, activating upon detection.
Integrated Face Detection technology with PPE Detection, minimizing the risk of injury and fatalities by identifying personnel and enforcing safety protocols.
Constructed a vehicle tracking model deployed on NVIDIA Jetson Nano, optimizing the model to INT8 precision for efficient performance.
Crafted a multi-stage Vehicle Type → Number Plate → Number Detection model, aiding the government in generating substantial revenue through penalties for traffic rule violators.

09/2021 – 06/2022

Developed Ethical AI models incorporating explanation mechanisms for prediction reasoning. Applied in Object Detection and NLP models.
Spearheaded various NLP Test Augmentation methodologies encompassing PII, Text Augmentations, and Counterfactuals among others.
Established an End-to-End Recommendation System Pipeline integrating content-based and collaborative filtering approaches.

02/2021 – 09/2021

Engineered an automated stock trading agent employing reinforcement learning for buying and selling actions.
Utilized RNN and LSTMs to construct several predictive models for forecasting future stock prices.

02/2021 – 06/2021

Worked on multiple Object Detection and Image Processing algorithms with more than a couple of clients.

03/2020 – 08/2020

Created a small Blogging website using JSP and Servlets, where senior authorities can share their orders, suggestions, and ideas with their juniors. Handed over the project to lead.

Skills