Laxmaan Balaji Email and Phone Number

Leveraged LLMs (GPT) to implement an AI Assistant that can interface with databases, existing backend APIs to answer questions, run backend functions and call APIs (internal or external), process and simplify the returned results. Designed a system in partnership with Confluent to serve as an airline booking assistant.Implemented a multi-language OCR model to log and parse receiptsand boarding passes during business travel

Involved in software development of various flavors ( front-end, backend,data science / ML) across different internal teams within Meta(Whatsapp, Instagram, FAIR, Reality Labs, Internal Tools)Committed a total of 1000+ significant lines of code to productionduring bootcamp phase across products such as facebook.com, FB app(android), Instagram, and Whatsapp.

Research Assistant on the DARPA SCORE project. The project involved designing a machine learning system that predicts if the results of a research paper are reproducible or not. Our feature extraction pipeline extracts the text and results from the research paper and our ML system - a synthetic prediction market uses the features to predict the reproducibility of the paper. Our system resulted in a publication in AAAI - 22.My role in the project involved designing and improving the… Show more Research Assistant on the DARPA SCORE project. The project involved designing a machine learning system that predicts if the results of a research paper are reproducible or not. Our feature extraction pipeline extracts the text and results from the research paper and our ML system - a synthetic prediction market uses the features to predict the reproducibility of the paper. Our system resulted in a publication in AAAI - 22.My role in the project involved designing and improving the prediction market. I was heavily involved in the theoretical as well as the programming aspect. In addition, I collaborated with the feature extraction team on NLP tasks such as summarization and extracting meaningful embeddings. Furthermore, I carried out software development tasks such as implementing version control and enabling basic CI/CD features for our system. I also created dashboards to display the results of the experiments and automatically generate graphs. Show less

TA for Discrete Mathematics course covering Boolean and Predicate Logic, Combinatorics, Cryptography Basics, Proofs, and Graph Theory.My responsibility was to hold recitations where I went over problems and supplemented the teaching done in regular sessions. Additionally, I held office hours.

TA for Numerical Analysis II. Numerical Analysis II presents a broad overview of numerical methods for solving major problems in scientific computing, including linear least squares, eigenvalues, optimization, boundary value problems for ordinary differential equations, partial differential equations, sparse linear systems, Fast Fourier Transform, and random numbers and simulation.

Worked on Information Extraction and Data Analysis. Research entailed identification of the impact of COVID-19 on Forex trading and analysis of top traders’ trading history for viable trends.

It is a graduate-level course introducing Machine Learning concepts, with programming exercises and projects in MATLAB.Responsibilities include aiding in dissemination of course material, grading projects and homework, and holding office hours

Learning Assistant for CMPSC 458 - Fundamentals of Computer Graphics.Responsible for the dissemination of course material, grading projects and assignments and holding office hours to clarify doubts the students enrolled in the course have.

Developed and debugged React web app components in an Agile environment.Additionally, designed an ML model for automated verification of healthcare professionals’ shift timings in Python with 96% accuracy. Achieved potential savings of up to 500 man-hours per month in verification.

Generated tailored responses to customer complaint emails by classifying the complaint’s domain using Word2Vec and CNNs implemented on Python. Achieved a test classification accuracy 0f 87.6%

Used JavaScript, AJAX, and Apache Tomcat to modify the modules in Fast ForWord™. Successfully adapted the modules in the program to Indian contexts.