AboutExperienceProjectsSkillsContactResume

Animish
Jain

Computer Science student at Cornell interested in applied AI/ML, quantitative software development, and software engineering.

ResumeGitHubLinkedInContact
Portrait of Animish Jain

About

About me

I'm a Computer Science student at Cornell University pursuing both a B.S. and M.Eng. in Computer Science through Cornell's Early M.Eng. (3+1) program. My interests lie at the intersection of machine learning, applied AI, and quantitative systems.

Through internships at Pfizer, Anote AI, and GEICO, as well as projects spanning NLP, computer vision, ML infrastructure, and financial modeling, I've developed a passion for building systems that create measurable impact. I'm currently seeking opportunities in software engineering, machine learning engineering, applied AI, and quantitative technology.

Experience

Relevant Experience

GEICO

Machine Learning Engineering Intern

Jun 2026 - Aug 2026Palo Alto, CA
  • Building a retrieval-augmented memory pipeline for LLM agents in Python using Hugging Face sentence transformers (BGE, MiniLM), dense vector embeddings, and cosine-similarity search to retrieve semantically similar historical execution traces during inference, enabling adaptive context injection for improved agent reasoning
  • Developing a modular case library framework in Python (curator, store, retriever, HTTP agent wrapper) integrated with LangSmith tracing and offline evaluation pipelines for scalable agent experimentation and rapid iteration
  • Implemented embedding pre-computation, JSON-based case schemas, and sub-100 ms semantic retrieval supporting 5,000+ historical cases while enabling automated A/B and regression testing with <2% performance degradation
PythonHugging FaceLangSmithSQLLangChainTrussed

Anote

Machine Learning Engineering Intern

May 2025 - Aug 2025New York, NY
  • Benchmarked zero-shot object-detection pipelines using YOLOv8, DINO, and Faster R-CNN, implementing workflows in Python with PyTorch and Ultralytics to compare mAP, IoU, and other ML metrics across models
  • Designed and deployed a smart orchestrator reasoning system using Python, Docker, and Ollama; engineered multi-agent workflows with LangChain and custom routing logic, reducing LLM inference latency by ∼25%
  • Integrated MCP into LLM pipeline using Python, FastAPI, JSON Schema, and LangChain, expanding agent tool interoperability by 40% and enabling tool-augmented reasoning across internal and external system endpoints
PythonPyTorchUltralyticsDockerOllamaLangChainFastAPI

Pfizer Inc.

Software Engineering Intern

Sep 2023 - Feb 2024Peapack, NJ
  • Queried over 50,000 internal license-request records using Splunk and SQL for downstream model training
  • Developed a multimodal language sentiment analysis algorithm in Python using TensorFlow and Scikit-learn with 5 individual machine learning models to evaluate the justification behind license and permission requests
  • Applied optimization techniques including L1/L2 penalties, dropout regularization, and learning rate scheduling to improve model performance, achieving 97.6% accuracy for automated request approvals and denials
PythonTensorFlowScikit-learnSplunkSQL

Featured Projects

Selected Projects

A collection of software, machine learning, and applied AI projects I’ve built.

Jun 2026

Texas Hold'em Monte Carlo CFR+ Poker Solver

  • Implemented a Python/FastAPI Monte Carlo poker simulation engine supporting up to 50,000 rollouts per evaluation, estimating Hero/Villain equity, win/tie/loss rates, action EV, return standard deviation, and 95% confidence intervals across sampled opponent ranges, board runouts, and strategic decision scenarios
  • Developed an abstraction-based CFR+ solver in Python configurable up to 100,000 iterations, using regret-matching+, positive regret clipping, average strategy accumulation, cumulative regret tracking, strategy stability metrics, and exploitability proxy reporting for convergence analysis over time
  • Built a full-stack heads-up NLHE solver over 1,326 possible two-card combos, with 52-card deck generation, 5/6/7-card hand evaluation, inferred Villain range matrices, and hand/board/SPR abstraction buckets
PythonFastAPIReactTypeScriptVite
GitHub

Apr 2026

Equity Options Volatility & Risk Engine

  • Built a vectorized SPY options pricing engine for cross-model comparative analysis implementing Black–Scholes, Binomial Tree, and Heston models, evaluated on 6.6M real contracts across 757 trading sessions from 2019–2021
  • Conducted a large-scale out-of-sample implied volatility surface study with liquidity filtering on 548K+ SPY options; Found that binomial tree achieved best model with 44.8% of predictions within ±10% of observed prices
  • Analyzed model robustness under calm vs. high volatility regimes (March 2020), finding Black–Scholes and binomial accuracy improved by ∼15–17% under high volatility while Heston declined ∼2%
PythonNumPyPandasSciPy
GitHub

Apr 2026

Event-Driven Backtesting Engine

  • Built a multi-asset event-driven backtesting engine that replays historical OHLCV bars through modular data, strategy, portfolio, execution, risk, analytics, and reporting components
  • Implemented train/test strategy evaluation over historical market data, comparing moving-average, mean-reversion, linear-regression, and logistic-regression strategies on out-of-sample performance
  • Engineered risk controls including per-symbol exposure caps, gross leverage limits, stop-loss exits, and max-drawdown portfolio de-risking to make strategy simulations more realistic
PythonPandasNumPy
GitHub

Oct 2025

Timbr

  • Developed a full-stack “Tinder-for-Homes” iOS app using swift, enabling users to swipe through real-estate listings, integrating real-time data and dynamic filtering to give users personalized property recommendations
  • Engineered responsive UI/UX components and backend APIs (React, Flask/FastAPI, PostgreSQL) to support seamless user onboarding, secure authentication, scalable data handling, and accurate prediction models
  • Built scikit-learn inference pipelines that featurized swipe directionality, dwell-time distributions, response latency, listing metadata, and user interaction sequences into calibrated ranking models for low-latency property preference prediction
SwiftReactFastAPIPostgreSQLScikit-learn
GitHub

Nov 2025

GreenTab

  • Built a full-stack sustainability analytics Chrome extension using JavaScript, Manifest V3, and Chrome APIs to monitor real-time browsing activity and estimate the environmental impact of spending time on the internet
  • Integrated backend APIs to calculate CO₂, energy, and water waste metrics via precise calculation models
  • Implemented user authentication using Google OAuth and persistent data storage with Supabase (PostgreSQL + SQL queries) to manage user profiles and browsing-impact telemetry to create personalized dashboards for users
PythonJavaScriptFirebaseChrome APIsSupabase
GitHub

Sep 2024 - Jan 2025

Skin Cancer Classification App

  • Evaluated multiple machine learning pipelines in Python using TensorFlow, Scikit-Learn, and OpenCV, comparing metrics across CNN, KNN, Logistic Regression, Linear Regression, and SVM models for skin cancer classification
  • Developed a multiplicative-weights update ensemble algorithm to optimally combine predictions from heterogeneous classifiers, improving overall accuracy, precision, and robustness against model-specific bias by over 10%
  • Built and deployed an iOS application using SwiftUI, integrating the trained ensemble model pipeline to allow real-time inference on mobile devices with a clean and accessible UI supporting seamless on-device classification
PythonTensorFlowScikit-LearnOpenCVSwiftUI
PaperPoster

Sep 2023

LetMeCook

  • Built a computer vision-powered recipe generation web app that lets users scan their fridge or cupboard and receive personalized recipes based on detected ingredients and dietary preferences.
  • Integrated Google Cloud Vision API to identify food items from user-uploaded images, then connected ingredient data with Edamam and OpenAI APIs to generate recipes with nutritional information and step-by-step instructions.
  • Developed responsive frontend workflows for image capture, photo editing, dietary preference input, and recipe display using React, TypeScript, Tailwind, Radix UI, and Lucide.
  • Helped implement food detection and recognition features while supporting demo/video materials for a hackathon project that won Best High School Hack at MediHacks 2023.
ReactTypeScriptTailwindFlask
GitHubDevpost

June 2024

JigTurn

  • Engineered an Arduino-based autonomous bicycle turn signal that lets riders signal turns without removing their hands from the handlebars.
  • Developed embedded software using IMU sensor readings over I2C to detect turn completion and automatically deactivate blinking LED signals.
  • Soldered and integrated Arduino Uno, IMU, battery pack, buttons, and LED panels into a custom bike-mounted prototype.
  • Led project management, testing, documentation, and user instructions for a three-person engineering team.
  • Tested with 12 riders, achieving 100% signal recognition from 50 feet and an 8.7/10 average user satisfaction rating.
ArduinoEmbedded SystemsI2CIMUHardware Prototyping
Engineering Report

Skills

Technical Skills

Languages

PythonJavaOCamlJavaScriptTypeScriptSQL

ML / Data

PyTorchTensorFlowKerasscikit-learnPandasNumPy

Web / Backend

ReactNext.jsTailwind CSSFastAPISupabaseDocker

Tools

GitHubAWSLinuxJupyterVS Code

Contact

Feel free to reach out!

Email[email protected]LinkedInlinkedin.com/in/animishjainGitHubgithub.com/AniJ15ResumeAnimishJainResume.pdf