About Achievements Experience Projects Skills Contact
Open to internships & research roles

Anish Lotake

BS Computer Science · UT Austin · Class of 2027

Top 0.8%
IMC Prosperity 3 · #102 Globally
Top 1.8%
JEE Main · 1.5M+ candidates
INMO
Math Olympiad Qualifier
14+
GitHub Projects

About Me

The Person Behind the Code

Anish Lotake
📍 Austin, TX

I'm a rising senior studying Computer Science at UT Austin, where I split my time between building production AI systems and pushing the boundaries of machine learning research.

My work spans the full stack — from Transformer architectures and RAG pipelines to Kubernetes deployments and Reddit-integrated apps. I'm drawn to problems at the intersection of math, finance, and engineering, which led me to place top 0.8% globally in IMC's algorithmic trading competition out of 12,620 teams.

Before Austin, I was solving math olympiad problems for fun — qualifying for the Indian National Math Olympiad and placing in the top 2% on India's most competitive entrance exams. That same precision and drive is what I bring to every project I build.

University
UT Austin
Degree
BS Computer Science '27
Focus Areas
AI/ML · Full-Stack
Location
Austin, TX

Awards & Recognition

Competing at the Highest Level

From national mathematics olympiads to global algorithmic trading — a track record of elite performance.

🏆
Top 0.8%
IMC Prosperity 3 — Algo Trading
#102 globally · #30 in the US · 12,620 competing teams worldwide
📐
Top 1.8%
JEE Main — Mathematics
Among 1.5 million+ candidates in India's largest engineering entrance exam
🔢
Top 2%
JEE Advanced — Mathematics
Among 180K+ candidates — India's most selective engineering examination
🎯
INMO
Indian National Math Olympiad
India's USAMO equivalent — national-level qualifier, top mathematical talent

Work & Research

Where I've Contributed

Production engineering and ML research at UT Austin — building real systems, not toy projects.

Feb 2026 – Apr 2026
Full-Stack Software Developer — SubredditPro
Center for Media Engagement · University of Texas at Austin
  • Built a Reddit-integrated full-stack application using TypeScript, React, and Flask with Google Cloud Firestore as the backend database
  • Designed backend APIs to query and filter large-scale Reddit datasets, transforming raw comment data into structured decision tasks for dynamic UI consumption
  • Implemented a data processing pipeline to sample, shuffle, and serve subreddit-specific content — enabling low-latency retrieval and consistent user experience
  • Embedded a full web application within Reddit's platform using Devvit, navigating strict platform constraints to deliver native-feeling UX
TypeScriptReactFlask Google CloudFirestoreDevvitReddit API
Jun 2025 – Aug 2025
ML Undergraduate Research Assistant
Dept. of Information, Risk & Operations Management · UT Austin
  • Engineered a Transformer-based architecture for portfolio optimization in TensorFlow/Keras — 4 stacked attention layers, residual connections, layer normalization, and dropout regularization
  • Collected, cleaned, and processed 60 years of financial data (1962–2022) from CRSP/Wharton Research Data Services, converting to efficient .parquet format
  • Designed a two-stream input structure combining factor exposures and realized returns, enabling dynamic portfolio weight allocation
  • Trained over 500 epochs with Adam optimizer — achieving validation loss ≈ 0.5, demonstrating stable out-of-sample generalization
TensorFlowKerasTransformer PythonPandasParquetFinancial ML
Graduating May 2027
🤘
The University of Texas at Austin BS in Computer Science  ·  Austin, TX

Projects

Things I've Built

Production deployments, AI systems, and cloud-native infrastructure — end to end.

MedicalGPT — Full-Stack AI RAG Assistant
Production RAG system with a decoupled FastAPI backend on Render and Streamlit client on Streamlit Cloud exposing 6 REST endpoints. Full pipeline: PDF ingestion → 800-token chunking → 3,072-dim embeddings (Google Gemini) → Pinecone serverless vector index → Groq LLaMA-3.3-70b answers with document + page-level citations.
✓ 100% factual-recall pass rate  ·  Cold start: 45 s → <1 s via GitHub Actions cron
FastAPIPineconeLLaMA 3.3 Gemini EmbeddingsStreamlitRenderRAG
Production Microservices on Kubernetes
Cloud-native voting app — React frontend → Go API → 3-node MongoDB replica set — hardened into a production deployment: resource limits, HPA autoscaling, PodDisruptionBudgets, default-deny NetworkPolicies, and an Ingress. Packaged as raw manifests, Kustomize overlays, and a Helm chart, with Docker Compose for local dev and a GitHub Actions CI pipeline that helm-lints and schema-validates every manifest on each push. Validated on a real cluster (kind). The live demo lets you scale pods, kill them to watch self-healing, and trigger autoscaling.
✓ Interactive cluster demo · Helm + Kustomize · HPA · self-healing · green CI/CD · validated on a real cluster
KubernetesHelmKustomize DockerMongoDBGoCI/CD
SQL E-Commerce Analytics — PostgreSQL
End-to-end analytics on a simulated online retailer: a normalized PostgreSQL schema, a reproducible synthetic dataset (~5K customers, ~57K order lines over 3 years), and 10 analytical queries — RFM segmentation, cohort retention, churn, lifetime value, Pareto/ABC, and a recursive category tree. The live dashboard charts every result and lets you run SQL against a real Postgres database running in your browser.
✓ Interactive Dashboard + In-Browser Postgres (WASM)  ·  Window Functions · Recursive CTEs · PL/pgSQL
PostgreSQLSQLWindow Functions CTEsData AnalysisRFM
Coffee Shop Android Application
Kotlin-based Android app for a coffee shop platform with scalable UI components using RecyclerView (grid & horizontal layouts). MVVM architecture with ViewModel manages asynchronous data flows and UI state, improving separation of concerns. Features category browsing, product listings, cart navigation, and Glide-powered image loading with loading states.
✓ MVVM architecture · Glide image loading · Android SDK best practices
KotlinAndroidMVVM RecyclerViewViewModelGlide
📈
Algorithmic Market-Making
Quantitative market-making strategies built for IMC's Prosperity 3 global trading competition. Covers pricing models, inventory management, and order book dynamics — the same codebase that put me in the top 0.8% globally (#102/12,620 teams, #30 in the US).
✓ #102 globally  ·  #30 in the United States
PythonJupyterQuant FinanceAlgo Trading
ShopE — E-Commerce Platform
Full e-commerce web application deployed live on Vercel. Responsive product catalog, shopping cart, and modern UI — built with clean HTML/CSS/JS and shipped to production.
✓ Live on Vercel
HTMLCSSJavaScriptVercel

Tech Stack

Skills & Technologies

Languages
Python TypeScript JavaScript C# SQL Java
AI / ML
TensorFlow Keras Transformer Architectures RAG Systems LLM APIs Pinecone Groq Google Gemini
Frontend & Backend
React FastAPI Flask Streamlit REST APIs Devvit
Cloud & DevOps
Kubernetes Helm Kustomize Docker Google Cloud AWS Git GitHub Actions Render Firestore PostgreSQL MongoDB Redis

Life Outside the Terminal

When I'm Not Working

💃
Latin Dance

Bachata is my favorite — there's something about the rhythm, connection, and musicality that's completely addictive. I love how it blends culture, movement, and expression into something you feel rather than just perform.

🎸
Guitar

Playing guitar is how I decompress. Whether I'm fingerpicking a slow melody or strumming through a bachata rhythm, music keeps me grounded when code gets complicated.

Contact

Let's Connect

Open to SWE internships, ML research roles, and conversations about AI, trading, or hard math problems.