Exploring the space where systems think and biology inspires.

From satellite observationscheduling at Leidos to LLM pipelines for cancer biology.

CS + Biology at UW-Madison | Computing for biology's next frontier.

GitHub

Featured Project

Architected failure-driven self-improvement loop for LLM agents across a 5,000+ task benchmark, outperforming ReAct baseline by 15pp via RAG-enhanced retrieval and Transformer neural network reasoning; built AgentBench + LLM-as-judge eval infrastructure with GitHub Actions CI/CD. Optimized inference via fine-tuning, task horizon tuning, and parameter search; built provider-agnostic serving abstraction supporting end-to-end ML pipeline scaled to horizon 20. Engineered LLM-as-judge harness scoring 7 dimensions; implemented novel repeated_failure_rate metric to quantify self-improvement efficacy.

PythonLLM APIsRAGTransformersAgentBenchInference OptimizationGitHub ActionsCI/CD

View on GitHub ↗

Self-ImprovingLLM Agent

00 / About

Hi, I'm Akash.

I'm a CS + Biology student at UW-Madison, graduating in May 2028. Right now I'm building LLM agent pipelines for spatial transcriptomics research at the Kendziorski Lab, and my industry experience has been in backend infrastructure and distributed systems at Leidos.

I grew up in Princeton, NJ, where I spent a lot of time on tennis courts (still do). I've coached players, organized tournaments, and built a racket stringing business out of it. That's where I learned that most interesting problems look messier up close than from a distance, and that knowing the fundamentals is what lets you improvise when things go sideways.

I'm drawn to work where the technical depth actually matters. Not as a marker of seniority, but because the problem just doesn't yield without it.

The goal is to build things that couldn't exist without both solid engineering and thoughtful design, where getting the craft right is the only way through. That intersection is where the most interesting problems are, and where I want to spend my career.

3.9 GPA · Dean's List · CS + Biology, UW-Madison · Graduating May 2028 · Open to Summer 2027 internships & co-ops

00 / Education

University of Wisconsin–Madison

B.S. Computer Science & Biology · August 2024 — May 2028

3.9 GPADean's List: Fall 2024, Spring 2025, Fall 2025

Computer Science

Data Structures

Advanced Algorithms

Deep Learning

Machine Learning

Artificial Intelligence

Linear Algebra

Machine Organization and Programming

Intro to Computer Engineering

Operating Systems

Database Management Systems

Cryptography

Differential Equations

Discrete Mathematics

Biology

Bioinformatics

Computational Biology

Genetics

Cell Biology

South Brunswick High School

September 2020 — June 2024

Computer Science

Computer Science Capstone

AP Computer Science A

AP Computer Science Principles

Android Application Development

Virtual Reality

Game Development

Web Development

Mathematics & Science

AP Statistics

AP Biology

Hover over any course to explore

00 / Experience

Where I've worked

Systems & Infrastructure

Leidos

Software Engineer Intern

May 2025 — Present

Optimized multi-objective satellite scheduling with memoization and dynamic programming in Python, then rewrote the engine in Rust for the team's production microservice ecosystem.

Designed a Kafka event-driven pipeline persisting to MongoDB with at-least-once delivery, decoupling mission-critical satellite operations.

Containerized a Docker test suite on Kubernetes via GitHub Actions CI/CD, with pytest coverage for boundary conditions and fault injection.

Built a Claude Code MCP plugin on an institutional knowledge graph mapping team ontologies and service dependencies, serving 30 engineers across a multi-team defense program.

PythonKafkaKubernetesDockerMongoDBDynamic ProgrammingGitHub ActionsCI/CDRustNeo4j

Want to know more about these? See the Skills section ↗

Research & Biotech

The Kendziorski Lab, UW-Madison

Student Research Intern

September 2025 — Present

Built ML pipeline for scRNA-seq, NGS, and spatial transcriptomics; agentic framework selects models across scVI/scANVI/scGen/AmortizedLDA, cutting GPU runtime 300% (80→27 min).

Built TransferAgent, automating NLP literature synthesis and validity scoring across ~500 papers, cutting review time ~60%.

Co-authoring a clinical genomics publication on model performance and pipeline reproducibility for glioblastoma gene therapy research, 10-person interdisciplinary lab.

PythonRLLM APIsScanpySeuratCARDscRNA-seqscVIscANVIWeights & BiasesGitHub Actions

Want to know more about these? See the Skills section ↗

Research & Biotech

Inspirit AI

Machine Learning Research Intern

Sep 2022 — Mar 2023

Built skin cancer detection pipeline (OpenCV + scikit-learn, NumPy/Pandas) achieving 95% classification accuracy on ~70,000 dermoscopic images; published as first author on 'Early Skin Cancer Detection Improvement' through Inspirit AI.

Trained and validated Random Forest classifier on miRNA sequences achieving 95% predictive accuracy (p < 0.05); validated via stratified k-fold cross-validation for cancer cell likelihood prediction.

Pythonscikit-learnOpenCVDecision TreeKNNML Research

Want to know more about these? See the Skills section ↗

Leadership

Princeton Racket Club

Tennis Coach & Tournament Director

May 2024 — Aug 2024

Directed 8 regional tournaments (380+ match entries, 100% on-time initialization) while coaching 3 classes of 8-10 athletes ages 8-65, in a dual tournament director and head coach role.

Guided 3 junior athletes to 150+ regional ranking points and 5 adults up a full 0.5 NTRP competitive level through data-driven, skill-adapted coaching regimens.

LeadershipCommunicationEvent Management

Want to know more about these? See the Skills section ↗

Leadership

Tennis Racket Stringing Services

Founder

Jan 2019 — Present

Founded and scaled an independent racket stringing business to 45+ clients, servicing 12-20 rackets monthly with a premium 2-day turnaround tier alongside the standard 3-4 day SLA.

Grew the client base through multi-channel grassroots marketing: word-of-mouth referrals, flyers at local parks, high school team referral discounts, and an Instagram page of stringing craftsmanship.

EntrepreneurshipOperationsClient RelationsInventory Management

Want to know more about these? See the Skills section ↗

01 / 05

00 / Projects

Things I've built

Self-Improving LLM Agent

Failure-driven agent that learns from its own mistakes. Tested on 5,000+ tasks, consistently outperforming ReAct by 15 percentage points.

Python / LLM APIs / RAG / Transformers· Skills ↗Click to explore ↗GitHub ↗

Lotus Health

Financial risk engine over a 1,080-node ICD-10 comorbidity graph, querying 45 million insurance records to compute 5-year disease cost paths.

Python / SQL / RAG / ICD-10· Skills ↗Click to explore ↗GitHub ↗

LLM Agent Pipeline for Spatial Transcriptomics

End-to-end ML pipeline for scRNA-seq, NGS, and spatial transcriptomics. Agentic framework selects and evaluates models across scVI/scANVI/scGen/AmortizedLDA; cut GPU computation 300%.

Python / PyTorch / LLM APIs / Scanpy· Skills ↗Click to explore ↗Research project, no public repo

SAT Practice Test Generator

GPT-4o tutor that finds your weak SAT domains and generates exam-matched practice questions on demand.

TypeScript / React.js / Node.js / FastAPI· Skills ↗Click to explore ↗GitHub ↗

Computer Vision Hand Tracking System

Real-time hand gesture and joint angle tracking pipeline using OpenCV and MediaPipe. Outputs biomechanical data for clinical motion capture, rehabilitation, and sports performance analysis.

Python / OpenCV / MediaPipe / Computer Vision· Skills ↗Click to explore ↗GitHub ↗

Ecological Conservation Game

Multi-threaded C++ simulation where AI bots compete for survival under configurable ecological constraints.

C++ / Raylib / Game AI / Multi-Agent· Skills ↗Click to explore ↗GitHub ↗

USTA Tournament Explorer

Android app for browsing and filtering USTA tennis tournaments by region, level, and surface.

Android SDK / Java / GraphQL / Mobile· Skills ↗Click to explore ↗GitHub ↗

microRNA Cancer Detection (miRcore)

Random forest classifier on microRNA expression profiles to predict cancer biomarkers. 95% accuracy (p < 0.05).

R / Bioconductor / Random Forest / microRNA· Skills ↗Click to explore ↗Research project, no public repo

00 / Skills

Always learning.

Here's where I am right now. Most of it I picked up quickly on real projects, and the list keeps growing.

Languages

Python·TypeScript·R·Java·JavaScript·C++·Rust·SQL·Node.js

Infrastructure

Docker·GitHub Actions·REST APIs·Kafka·Kubernetes·MongoDB·FastAPI·Redis·DuckDB·Neo4j·System Design

ML & Data

Machine Learning·PyTorch·LLM APIs·NumPy·Pandas·Scikit-learn·Weights & Biases·TensorFlow·OpenCV·MediaPipe·Whisper·CUDA·Decision Trees·KNN·Cross-Validation·Statistical Modeling

Agentic & LLM

RAG·LLM Agents·Semantic Similarity Search·Vector Databases·AgentBench / Benchmarking·Inference Optimization

Bioinformatics & Genomics

Scanpy·Seurat·CARD·Bioconductor·scVI·scANVI·scGen·AmortizedLDA·NGS·Spatial Transcriptomics

Frontend & Mobile

React.js·Android SDK·GraphQL·Cytoscape.js·Deepgram·Raylib

Developer Tools

GitHub·pytest·Code Review·OOP·Claude Code / MCP

Leadership & Product

Leadership·Communication·Event Management·Entrepreneurship·Operations·Client Relations·Inventory Management·Product Design·Spatial Computing·AR (Augmented Reality)·Accessibility

Hover over any skill to explore

00 / Accomplishments

Recognition along the way.

2nd Place, 2-Person TeamJul 2026

TEL: Spatial Intelligence Ideathon

UW-Madison

—Placed 2nd at a competitive UW-Madison ideathon, pitching a dual-product airport accessibility platform combining ZaiNAr sub-meter indoor positioning with SGA AR spatial computing.
—Designed a B2B tool giving airline staff real-time sub-meter wheelchair-passenger tracking between gates, paired with a B2C app letting family track accessible relatives through the terminal.

Spatial ComputingARAccessibilityProduct Design

Want to know more about these? See the Skills section ↗

00 / Contact

Get in touch

Open to Summer 2027 internships and co-ops in software engineering, AI/ML, data science, biotech, and computational biology research. Available May 15 – Sep 1. Email is the best way to reach me.

UW–Madison

GitHub

AkashGogate

akashgogate

Akash Gogate — 2026Built with Next.js · Framer Motion · Tailwind CSS