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

I was promoted to Software Engineer - ML based on all my successful deliverables in just one year. I am helping to make LLMs more accessible in Asset Management BU.

Contributions :

• Developed AutoTLM, a fine-tuned RAG model specialised in helping in TLM upgrade activities. For ex. Java 8 to Java 17, Python 2.7 to 3 or any tech stack upgrade of any application. Thus helping developer reduce 70% of their effort on such tasks.
• Implemented an end-to-end API performance suite mainly using JMeter, Jenkins, and Datadog to test all LLMs exposed in our Reverse Proxy Application, which also mitigated all EKS deployment bottlenecks present.
• Developed an LLM Reverse Proxy Application, enabling users to integrate LLM's capabilities in their individual apps securely. I also configured it's CI/CD and deployed it in EKS environment till Production.
• Created a RAG model for Q&A over any large codebase or text corpuses	.
• Increased unit test code coverage from 0 to 95% for Python based Proxy Application using pytest.
• Developed an automated Jenkins CD that reduced EKS deployment time from 35 mins to 9 mins.

Kenneth Warren & Son

Wells Fargo

Keyes Capital

Burman Family Holdings

Oxeon Partners

Sterling Private Capital

Fidelityag provides an artificial intelligence-based solution for waste sorting. Its system uses Artificial Intelligence, Computer Vision, and Material Recognition techniques to differentiates recyclable and organic materials from waste. It also provides solutions for fertilization and pesticides.

Provider of automated waste sorting solution

Working in CPS (Common Platform Solutions) DataAPI team, under AM Tech Business Unit. Developing Rest APIs and Cloud-based solutions to bridge the gap between Common Data and its consumers.

Contributions :

• Mitigated End-Of-Life (EOL) dependencies in our Java application, resolved bugs, and made minor improvements based on consumer feedback.
• Improved code quality by eliminating critical and blocking issues identified in a Sonar Report, increasing code coverage from 60% to 90%.
• Implemented OAuth-based authentication using (SSO) for Snowflake-based application users.
• Successfully migrated an application from TCWS (Tomcat-Web-Services) to a SpringBoot pattern, including upgrading services written in Java, configuring CI/CD, and deploying to EKS. This transformation improved microservices compatibility, resiliency, and reduced deployment time from 30 minutes to 5 minutes.
• Completed POC of GraphQL support with a focus on filtering and stitching use cases in APIs.
• Added BLOB (Binary-Large-Object) data type support in our API application for read and write use cases, allowing multiple file attachments of any type.

Stretch/Additional Projects (AI/ML):

• Conducted a comparative study of Silverkite and Prophet Time-Series models on daily data. Replacing Prophet with Silverkite resulted in a 15% decrease in MAPE and a 17% decrease in RMSE, with consistently better results across different stock prices.
• Working on a POC to automate Total Lifecycle Management (TLM) activities, including Java 8 to Java 17 upgrades and TCWS to SpringBoot migrations.
• Developed a Binary Classification pipeline with 92% accuracy that assists Quant Portfolio Managers in selecting the most suitable optimisation model runs for portfolios, eliminating true negatives and improving scaling.
• Analysed the Volatility in the market with the performance of industries over time drawing insights, thus helping Portfolio Managers make informed decisions before making an optimisation.

• Completed Fidelity's Leap Program, learning various technologies related to Full Stack Development.

• Completely re-designed the existing custom grids in a portfolio management application, using AG-Grid.
• Reduced the time it takes to add new functionalities on the grid from 2-3 story points to a few minutes.
• Fixed issues like browser crashing in case of large volumes of data present in some portfolios.
• Performed UI automation testing using Cypress and scenario-based manual testing.

• Designed and implemented an end-to-end pipeline for time series forecasting, from raw data ingestion (updated daily) to forecasting for upcoming days in future.
• Constantly improved forecast methodologies using either exogenous variables, statistical, machine, deep learning models or implementing techniques of various research papers.
• Worked with over 2000+ real-world datasets in a single pipeline and continuously achieved well below 29% MEAN MAPE over various data domains.
• Reduced daily-forecast generation runtime by 50% through implementing parallelism in python and using a custom-automated pipeline.

Providing research consultancy in the areas of Machine learning, NLP, Financial Analytics and Signal Processing

• Developed a comparative study of various statistical, machine and deep learning methodologies namely SARIMA, Multi-variate Linear Regression, XGboost and Stacked LSTM for predicting the spread of Vector-borne diseases.
• Achieved 90% accuracy over RMSE and MAE metrics through implementing a robust EDA and using GridSearchCV. The study was based on New York and Los Angeles monthly infectious cases datasets.
• Compared the temperature projections from the World Climate Research Programme's CIMP3 multi-modal dataset with the actual decadal temperature variations of four cities around the globe and drew insights using data analysis.

University offering multiple courses and degrees across diversified streams

• Developed a dynamic dashboard primarily using Angular and Spring Boot services, increasing performance by 25% from previously used business insight and reporting tools by the Asset Management Team.
• Familiarized myself developing and consuming RestAPIs, handling CORS, using design principles, JDBC, etc.

1. Working on Research in the fields of Time Series analysis.

2. Developing personal projects related to Finance and Data Science.

The first and only student run multi-disciplinary lab of SRM University at Chennai and Amaravati. The lab has currently 400+ members and associates with it.

Students from 8 departments work in 6 research groups concentrating on Artificial Intelligence, IoT, Computational Biology, Electrical Systems, Extended Reality and Cryptocurrency. The lab follows no hierarchical structure and fosters student ideas and interests. Next Tech Lab has advisors from around the world from distinguished affiliates and alumni of institutions like MIT, University of Cambridge, Google Brain etc. 

Next Tech Lab is the Silver Winner of QS Reimagine Education Awards 2018 in Student Led Innovation domain awarded at the Wharton School, University of Pennsylvania.

SRM Next Tech Lab is a student-led R&D hub fostering collaboration and innovation across diverse engineering and science domains

Worked on five Research Papers under the guidance of Dr. Sarbani Roy, Professor at Jadavpur University, on the following topics :

1. Long Term Time Series Pollution Forecast using Statistical and Deep Learning Methods, based on Particulate Matter (PM2.5 and PM10)  data of Kolkata, India.

2. Improved Temporal Predictions through Time-Series Labeling using Matrix Profile and Motifs.

3. Proposed a novel Spatial-Temporal pollution forecasting model using a Hybrid Network, consisting of an ensemble of Deep Learning and Statistical approaches.

4. Presented a Hybrid Learning Model for Spatio-Temporal Forecasting of PM2.5 using Aerosol Optical Depth.

5.  Presented the impact of Lockdown on City Air Pollution in India: A Case Study for short-term environmental policy-making by statutory government bodies.

Jadavpur University is a public state university located in Kolkata, West Bengal, India.The National Institutional Ranking Framework has ranked it 12 among engineering institutes in India in 2018, 13 overall and 6th among universities. JU has achieved autonomous university status by UGC in 2018.

Jadavpur University is a Educational Institute, Research and Consultancy

1. Implemented a voting system based on Liquid Democracy using SmartContracts and led research on scalability of Blockchain in today’s world.

2. Mentored newcomers to get into the field of Blockchain.

1. Mentored newcomers on Blockchain Development.

2. Actively participated in Capture The Flag contests.

SDE-ML @Fidelity Investments | Data Science | Machine Learning | Deep Learning | GenAI | LLMs | Python | SQL | MLOps | Software Development

Pratik Saha

He is currently working as a Software Engineer - Machine Learning at Fidelity Investments, Bangalore, India and his published research papers have garnered more than 100+ citations till date.