About
Hey there! If you've stumbled upon this little corner of the internet with the intention of hiring me, congratulations on your excellent life choices! Stick around till the end because, surprise, I crafted this digital masterpiece just for you.
I'm Harshitha — an Applied Scientist who went from writing RISC-V assembly at IIT Madras to building agentic LLM platforms at Amazon Web Services. Along the way I fine-tuned hedge fund AI at Millennium Management, got awarded Intern of the Month at Cisco (yes, framed), and somehow convinced a cancer-survival prediction model to hit 0.91 recall. I hold a Master's in Computer Science from NYU and I have strong opinions about RAG pipelines and mediocre opinions about sleep schedules.
My rapid learning ability is a cornerstone of my career — I thrive in alien environments, which is great because ML keeps inventing new ones every six months. I'm looking for opportunities that challenge me, value innovation, and ideally have good snacks.
Industry Experience
A track record of applying machine learning and AI to solve high-impact problems across industry and research settings.
Amazon Web Services, California
Oct 2024 – Present
Applied Scientist — Security Services & Observability
Built an agentic LLM platform using Strands SDK with custom tool-calling and state-transition logic to automatically map AWS-managed security controls to NIST, PCI, and CIS frameworks, eliminating manual regulatory workflows across multi-account environments. Led a team of 3 scientists developing SFT workflows with BERT-based encoder and cross-encoder models for compliance document classification and incident-severity prediction. Designed multi-stage retrieval and ranking pipelines, LLM observability frameworks via Langfuse, and a GraphRAG system using Amazon Neptune for automated compliance evaluation and audit-ready reasoning.
Millennium Management LLC, New York
Jan 2023 – Oct 2024
Applied Scientist — Central Data Platform
Built and deployed a Retrieval-Augmented Generation framework that parsed 35,000+ research documents, significantly improving search relevance and precision for firm-wide investment workflows. Fine-tuned LLaMA models using LoRA and developed BERT-based classifiers for industry-level text classification, automating the tagging of broker research documents. Served as subject matter expert and team lead (4 engineers) on the Broker Research initiative, driving architecture design, agile execution, and cross-functional collaboration with Product Managers to deliver production ML systems.
Jun 2022 – Aug 2022
Data Science / AI Intern
Automated SLA check generation using ML and Python OOP, creating 10,000+ data delivery checks to ensure timely pipeline execution. Migrated 10 broker research datasets to cloud infrastructure by enhancing the ETL framework with automated testing via Python, Jenkins, Git, and AWS.
Cisco Systems, Bangalore
Feb 2021 – Jul 2021
Software Engineer
Developed and productionized a machine-learning classification system for automated network issue triage, reducing incident resolution latency by 25%. Implemented end-to-end training, validation, and deployment pipelines for text-based issue classification using Python and classical ML techniques.
Palms Connect LLC, San Diego
Jun 2020 – Sep 2020
Analytics Intern
Predicted cancer patient survival outcomes using ensemble ML algorithms — Random Forest, Gradient Boosted Trees, and ANNs — achieving 0.91 recall on a highly imbalanced medical dataset. Conducted comprehensive feature analysis and communicated data insights through Matplotlib, Seaborn, and Tableau visualizations.
National University of Singapore, Singapore
Dec 2019 – Jan 2020
Machine Learning Intern
Completed a rigorous 5-week intensive program on Machine Learning, Deep Learning, and Reinforcement Learning led by NUS faculty and Hewlett Packard Enterprise, achieving a 90% overall grade. Engineered a real-time image classification inference application with a continuous deployment pipeline using Python's Flask framework.
Indian Institute of Technology Madras (IIT-M)
Nov 2018 – Jan 2019
Software & Research Intern
Developed a Flask-based front-end for automated configuration-driven testing on the RISC-V platform and migrated test codes to the production architecture. Collaborated with a team of 10 researchers, including members from UC Berkeley, on processor functionality testing using RISC-V assembly.
Skills
Things I'm good at. Some of them I'm annoyingly good at.
AI & Machine Learning
NLP & LLMs
Languages
Data & Cloud
Engineering & Tools
Visualization & BI
Certifications & Publications
19 publications and counting. Apparently I write papers the way some people write tweets — compulsively and often.
Certifications
Google Certified — TensorFlow Developer
TensorFlow model development across Computer Vision, CNNs, NLP, and real-world image data strategies.
AWS Certified Machine Learning — Specialty
Validates expertise in building, training, tuning, and deploying ML models on AWS.
AWS Certified Solutions Architect — Associate
Validates ability to design distributed systems and scalable cloud architectures on AWS.
Selected Publications
Deep Learning for Automated Detection of Lung Cancer from Medical Imaging Data
Automated lung cancer detection from chest X-rays and CT scans using deep learning. Published in IEEE (2023).
Temporal Analysis of Human Serum Albumin with RNNs for Changepoint Detection
RNN-based prediction of protein movement and changepoint detection in binding-affinity time series. Published in Springer (2021).
Ethical Considerations in AI-Assisted End-Of-Life Care Decision-Making
Examining the ethical implications of deploying AI in end-of-life healthcare decisions. Power System Technology (2023).
Security Challenges and Solutions in AI-Enhanced Cloud Platforms
Security challenges at the intersection of AI and cloud platforms, with proposed mitigation strategies. Power System Technology (2023).
Improved Cluster-Based Fault Tolerant Data Aggregation in Wireless Sensor Networks
TCP-based intra-cluster protocol minimizing energy consumption in WSNs. ICTACT Journal (2019).
Green IoT (G-IoT): An Insight on Green Computing for a Sustainable Future
Responsible application of IoT technologies toward energy efficiency and a greener future. Springer (2020).
Writing
I write about AI/ML on Substack — no hype, no fluff, just the stuff that actually matters. Free forever.
Not All Attention Is the Same: A Guide to Attention Mechanisms
A breakdown of bidirectional, causal, prefix, block, disentangled, and cross-attention — what they are, why they exist, and where each one shows up in modern language models.
Read →Attention Works Because of These Small Design Choices
The transformer's magic isn't just attention — it's the small decisions around it. Dimension selection, scaling before softmax, residual connections: the details that make training at scale actually work.
Read →Agentic Systems Are Not Magic, They Are Just State Machines with Better Marketing
AI agents demystified. Under all the hype, they're structured state machines using tool-calling, goal decomposition, and multi-agent coordination — not autonomous intelligence. Here's how they actually work.
Read →Testimonials
Contact
Feel free to reach out via email or LinkedIn.