https://dream.weekday.works/redirectToHiringCopilotExtension/

Managed the end-to-end process of training, testing, and deploying sentiment analysis and moderation services, incorporating advanced features like emotion detection and spam filtering. Enhanced moderation reliability and user experience through robust, scalable models.
Adapted large language models (LLMs) such as LLaMA3-8B, Phi3, LLaMA7B, and Mistral7B to specific tasks in Generalized Classification and text analysis. Utilized fine-tuning techniques like LoRA, QLoRA, and Supervised Fine-Tuning (SFT) to boost model accuracy and relevance in targeted applications.

Enhanced Instruction Adherence with Advanced Prompting Techniques:
Developed and applied prompting strategies—Chain of Thought (CoT), Branch of Thought (BoT), and Retrieval-Augmented Generation (RAG)—to address instruction-following challenges in LLMs. These innovations improved model consistency and responsiveness to user instructions.
Implemented optimization techniques such as pruning, quantization, efficient decoding, and prompt compression, reducing memory requirements by 60-70% and achieving a twofold increase in inference speed. This resulted in significant efficiency gains for high-demand applications.
Developed domain-specific datasets to fine-tune models, increasing instruction adherence and reducing hallucinations. This dataset curation improved model reliability and alignment with intended outcomes in real-world applications.:
Designed a summarization widget with charts and drill-down capabilities to provide quick insights into displayed data. Leveraged RAG and prompt engineering to ensure efficient and contextually accurate summaries, helping users access key insights seamlessly.

Hellman & Friedman

Sixth Street Growth

Temasek

Wellington Management

EDBI

Valor Capital Group

Battery Ventures

Suneight Investment

Intel Education Accelerator

Iconiq Capital

Plug and Play Tech Center

Azure Capital Partners

Recruit Strategic Partners

Hercules Capital

Top Tier Capital Partners

Sway Ventures

Glean Management

Globivest

SharesPost 100 Fund

Section Partners

Plataforma Capital

JC2 Ventures

Conductive Ventures

HTIF

Cloud based customer experience management software. The platform allows brands to turn public conversations on modern channels into valuable insights to improve marketing, advertising, customer service, product development, and more. Product features include location based analysis of brand experiences, content marketing, marketing analytics, reputation management, and more.

Cloud based platform offering customer experience management software

> Developed a full-scale video summarization tool utilizing frame selection, image captioning, and GPT-3.5 technologies.
>Constructed a large-scale retrieval system for searching similar queries by employing vector search techniques.
> Fine-Tuned large language models (LLMs) to build engaging conversational systems.

•	Built a python script that automated the movement of data from the company’s ERP system to its CRM system
•	This increased the order success rate from 60% to 95% and reduced avg. order placement time by 60%
•	Utilized ML algorithms such as Linear Regression & Radom Forests to find correlations between various variables
•	Built a python script which automated the process of price optimization with ML algorithm and updation in backend

BNP Paribas

MUFG

Synergy

Apollo Global Management

Manufacturer of cement. It also produces cementitious materials such as aggregates, additives, plaster, and concrete.

•	Selected by 800+ students as Hostel Legislator of Ganga Hostel for the year 2022-23 to represent the hostel in SLC
•	Responsible for regularly conducting the general body meetings and validating the hostel budget of 5.5 Lakhs
•	Actively participating in the debates of SLC on Issues in Institute and Chairperson of Ad-Hoc Committee on BSc-DS

• Worked in a team responsible for building the controller of prototype of a Biped Robot for Agricultural fields
• Learned about webot and G-mapping tools for simultaneous localization and Mapping for Path Planning
• Collected and labelled data for the tomato localization with labelimg in Yolo and pascal Voc Format 
• Implemented research papers for detecting and localization of tomatoes and peduncles and fine-tuned YoloV5, achieving mAP of 73.1

Indian Institute of Technology - Roorkee is among the foremost of institutes of national importance in higher technological education and in engineering, basic and applied research. Since its establishment, the Institute has played a vital role in providing the technical manpower and know-how to the country and in pursuit of research. The Institute ranks amongst the best technological institutions in the world and has contributed to all sectors of technological development. It has also been considered a trend-setter in the area of education and research in the field of science, technology, and engineering.

The Institute has completed 150th year of its existence in October 1996. On September 21, 2001, an Ordinance issued by the Government of India declared it as the nation's seventh Indian Institute of Technology. The Ordinance is now converted into an Act by the Parliament to make IIT, Roorkee as an "Institution of National Importance".

The Institute offers Bachelor's Degree courses in 10 disciplines of Engineering and Architecture and Postgraduate's Degree in 55 disciplines of Engineering, Applied Science, Architecture and planning. The Institute has facility for doctoral work in all Departments and Research Centres.

IIT Roorkee, is the oldest Technical Institution of Asia, and is devoted to the advancement of knowledge and education.

•	Built an Algorithm with which one can calculate the greenery score of an area with longitude, latitude, and radius 
•	Used open street map API to fetch the data for the algorithm and calculated the green score of Delhi to 24%  Blog
•	Analyzed NASA’s fire data from the MODIS satellite and built a fire count prediction system for a region Blog
•	The developed region-based fire prediction system has the potential to save $10s Mln for fire-affected companies

•	Analyzed hourly air composition data to find the correlation between concentrations of various gases present
•	Treated the columns with missing values with knn Imputation and dropping with a threshold of 30% missing values
•	Built a LSTM model and trained it on day wise gaseous composition data and achieved an average RMSE of 0.0012

FTS in association with “Janta Seva Mandal, Ballia Uttar Pradesh”, has been started in 2020 to serve people by normalising failures and uplifting an individual’s growth !!We are here to make people share not just their success stories,but also their failures.Each one of us can play a role in shifting how we collectively view failure by sharing the stories of how we failed. The more we talk about and hear stories of failure, the more at ease we will be about putting our toe in the water, raising our hand, trying that new thing and taking the leap. Stop keeping failures hidden. Let’s bring them out into the light and see them for what they really are…opportunities for learning and growing. What is your failure story? Are you ready to share?

FTS Group provides workplace training, industrial recruitment, warehousing and logistical solutions to transport industries.

•	Worked on building a Web App for monitoring the electric energy consumption of shops in a city mall
•	Developed a React web application using Django for the backend and ReactJS for the front end
•	The application had a 3-tier architecture with a primary dashboard for users, employers, and administration staff

• Responsible for managing the Tech Lab project under NSS, IITM, which aims at taking IITM’s resources to schools in remote 
locations for the empowerment of students through wholesome education.
• Filling the gap between theoretical and practical learning in students by providing local students exposure to STEM. 
• Leading a team of 50 student volunteers from IITM to embark curiosity in the minds of students and to wake up the engineers 
inside them.

•	Worked In a team of 30 developers responsible to design the backend with MVC architecture and documentation 
•	Built Session-based Authentication systems with the functionality of OTP and Personal Tokens using HTTP Protocol
•	Developed Various APIs for sending SMS and emails for backend of an App achieved 10k+ downloads on play store

Chillitray Technologies is a leading development & consulting services provider with clients across the globe. Primary areas of services include Web application development, App development, Cloud solutions, Automation solutions, E-Learning Products, AI/ML Modules, Cybersecurity solutions, and more. Our strength lies in understanding our client’s business processes, culture, vision, and goals across the industry segments and offering reliable client-oriented solutions. We believe in the digital revolution and have been supporting various starups in their MVPs.

Sniffing insights from the social data - A DataScientist @sprinklr|| Machine Learning Researcher

Abhishek pandir

A seasoned ML engineer with expertise in building end-to-end ML solutions—from the research phase through to deployment—plus experience developing websites and mobile apps. I’m currently working with Large Language Models (LLMs), pushing the boundaries of what AI can achieve. With a BTech from IIT Madras, I have a solid foundation in engineering principles and a talent for transforming complex ideas into impactful solutions.

Unpacking Granite 3.0’s Data Strategy: The Power of Curated, Public, and Synthetic Datasets in AI Model Development

Part 1 Granite 3.0:  The Data Powering Granite 3.0

Data compression using Principal Component Analysis (PCA)