Jaichandra J.

Tech Stack: Python, OpenAI, LangChain, Scrapy, AWS, PostgreSQL, FastAPI, Docker, GCP • Built a centralized database of licensable IP by scraping 160+ R1 university tech transfer portals, structuring data for semantic search and commercialization insights. • Integrated Grants.gov, SBIR, NSF, NIH, and SAM.gov APIs to enable automatic discovery of relevant R&D funding for startups and researchers. • Engineered a modular RAG pipeline with multi-step reasoning, query rewriting, and fallback handling to support robust, conversational Q&A over 35K+ IP documents. • Deployed production-ready FastAPI endpoints for semantic search and chat-based retrieval, forming the backend of an upcoming IP/funding discovery platform.

- Built FundLaunch AI Beta end-to-end, an intelligent system that automates investment fund generation from minimal user input by leveraging market research, LLMs, and 50+ workflow nodes – reducing fund creation time by 99% and transforming a multi-week manual process into a polished, investor-ready deliverable within minutes. - Developed intelligent AI chatbots to replace expensive consultations and research delays. Implemented context-aware conversations with query routing across research, coaching, and education domains. - Engineered a real-time streaming architecture using WebSockets for live progress tracking, instant feedback integration, and on-demand content regeneration and modification – transforming the user experience from static to interactive. - Automated investor prospecting and scoring, building a CRM pipeline that finds qualified investors per fund and a 7-level quantitative fund readiness algorithm, replacing subjective fundraising assessments with data and market driven accuracy. Tech Stack: Python, LangChain, LangGraph, LangSmith, AWS, Supabase, OpenAI, Perplexity, Docker, NextJs, React

Tech Stack: Python, FastAPI, LangChain, Docker, Groq, Assembly API, Streamlit, GCP • Developed an AI-powered Rubric Generator, simplifying grading by allowing educators to generate rubrics based on inputs like grade level, point scale, and assignment details, with PDF output, cutting down syllabus generation time up to 50%. • Built a Notes Generator that extracts concise, structured notes from sources like PDFs, PPTs, URLs, and YouTube videos, generating summaries and outlines in PDF format for efficient information processing. • Engineered a transcription and quiz tool to convert class recordings, YT videos and PDFs into text, providing users with summaries and automated interactive quizzes for better comprehension and engagement.

I am working as a Graduate Teaching Assistant for the course ITCS 6100 - Big Data Analytics for Competitive Advantage under Dr. Pamela Thompson. My responsibilities include grading assignments, conducting office hours and mentoring 80 students by providing hands-on training with AWS services such as S3, Kinesis, Glue, and Redshift to build and optimize data pipelines for analytics and machine learning applications.

I am working with Dr. Li Yang and the team for the NeurIPS 2024 Edge-Device LLM Competition, sponsored by Huawei, where we are tasked to build memory-efficient LLMs for mobile devices, reducing memory and energy consumption while optimizing performance in constrained edge environments. Our approach is as follows: • Utilized StreamingLLM to optimize long-sequence processing by retaining key tokens, achieving up to 22x speedup for continuous streaming on edge devices. • Integrated Wanda's pruning technique to reduce model size by eliminating non-essential weights, maintaining accuracy without retraining and enhancing energy efficiency for mobile deployment

I worked as a Teaching Assistant in ITCS 3162 - Introduction to Data Mining under Dr. Pamela Thompson where my responsibilities were to conduct office hours, grade homework and assignments, discuss project ideas with students etc.

I'm elated to announce that I'm starting my role as a Teaching Assistant for the course Database Systems, under Professor Dr. Sara Riazi. My responsibilities include conducting office hours, grading assignments and design challenging yet useful course work

Worked with the KL University's Research Team to Improve speaker recognition accuracy by 8% for numerous Indian languages, optimizing speaker modelling, feature extraction, and sequence modelling techniques to enhance performance. Collected and processed 100 hours of speaker data to improve model training and validation. The project, sponsored by the Ministry of Electronics and IT, India, is part of the Bhashini app, aimed at enabling cross-language communication using advanced speaker recognition technologies for Indian languages.