Founding Engineer | Applied AI Scientist

Minimal input, maximal output.

I build production AI systems that are practical, fast to ship, and resilient at scale. From AI voice reception to legal-domain LLMs, I focus on outcomes over hype.

Email GitHub LinkedIn
  • Phone260-498-4812
  • BaseUnited States
  • DegreeMS Computer Science, Purdue

Experience

Founding Engineer, AI Voice Platform

Better Collision Centers | Oct 2025 - Present

  • Built an omnichannel AI receptionist on Twilio SIP + n8n with retries, idempotency, and observability; improved CSI by 30%.
  • Built MVP and production backend for BetterCards in Flutter and Golang in 4 days; campaign brought in 1500+ clients in first run.
  • Led AWS architecture for BetterX AI services (RDS, DynamoDB, SageMaker/EMR, App Runner), including IAM hardening and monitoring.
  • Built a production scheduling optimizer with event-driven replanning and LLM feedback loops; reduced cycle time from 15 to 10 days.

Applied AI Scientist, Founding Team

New Territory Inc | Feb 2025 - Oct 2025

  • Designed a Mixture-of-Experts legal assistant model with gated top-k routing; achieved 1.7x token throughput and lower training cost.
  • Curated a 15T-token corpus via AWS Glue pipelines and Selenium crawlers with deduplication, PII filtering, and domain tagging.
  • Delivered a secure chatbot backend in 7 days with RBAC, role-aware RAG retrieval, chat sharing, and audit logs.

Research Assistant, Adversarial NLP

Purdue University | May 2024 - May 2025

  • Fine-tuned 50+ LLMs including BERT, ALBERT, LLaMA 3, and GPT-Neo for binary and multi-label classification tasks.
  • Automated robustness benchmarking with BAE, TextFooler, and TextBugger, producing reproducible attack and perturbation reports.

Software Engineer, Internal Platforms

Tata Consultancy Services (Client: CIBC) | Jul 2021 - Jul 2023

  • Built Java/Spring Boot gRPC backend for lending workflows, scaling to 7,500+ applications/day with low-latency services.
  • Implemented real-time decision rules that cut decision time from about 30 minutes to 30 seconds and increased auto-approval rates.
  • Improved risk checks and reduced post-decision overturns by 30% through stronger policy and validation pipelines.
  • Implemented Redis region-branch caching with TTL and event invalidation, increasing hit rate and reducing backend load.

Education

Purdue University

MS in Computer Science (Thesis) | Aug 2023 - Aug 2025

Fort Wayne, Indiana

Reading, Research, and Interests

System Design Learning

Papers, docs, and practical explainers

  • I regularly learn from YouTube system design channels, then validate design choices with production docs.
  • Core sources I revisit often: ChatGPT/OpenAI docs, Redis docs, PostgreSQL docs, and architecture blog posts.
  • Topics I track: caching strategy, indexing, consistency tradeoffs, queueing, retries, and distributed fault tolerance.

LLM Research

Applied and research-driven exploration

  • I actively research LLM systems, including model efficiency, retrieval-augmented generation, and evaluation quality.
  • I am especially interested in expert routing, inference optimization, safety constraints, and production reliability.

Theory Interest

Long-term scientific curiosity

  • I am deeply interested in Quantum Field Theory and enjoy reading conceptual as well as mathematically grounded material.
  • This interest shapes how I think about abstraction, modeling, and first-principles reasoning in engineering.

Technical Skills

Languages

Golang, Python, SQL, JavaScript, HTML/CSS

Frameworks

Flutter, React, Node.js, Tauri, gRPC

Cloud & Tools

AWS, Azure, Docker, Kubernetes, Git

Libraries

PyTorch, Hugging Face, Redis, Protobuf, pandas, NumPy, Selenium, CUDA