ML Engineer • Agentic Systems • LLM Apps

I build AI agents and data-backed products that ship.

3+ yrs across ML research, backend, and cloud. I design agentic workflows (LangGraph/LangChain), search & retrieval, and realtime pipelines—then deploy them with FastAPI, Docker, and cloud infra.

See projects Get in touch
GitHub LinkedIn Website
Now open to roles

Hi there, I’m

Krishna Vamsi Dhulipalla

ML/AI Engineer | Agentic Systems | RAG & Evals | Data Pipelines | Cloud-native MLOps

🎓 MEng CS — Virginia Tech 🤖 Generative AI & Agents (LangGraph/LangChain) 📦 Data Eng (Kafka • Spark • Airflow) ☁️ Docker • K8s • FastAPI 🧪 RAG • Observability

I ship production AI features end-to-end — from data ingestion and retrieval to evaluation, deployment, and monitoring — with a focus on speed, reliability, and clear success metrics.

  • Impact first: define latency/task-success goals and validate with A/Bs before scaling.
  • Reliable agents: hybrid retrieval (BM25+vector), tool-use guardrails, and regression evals to prevent drift.
Profile photo

Interactive 3D tilt on hover; motion is disabled if the user prefers reduced motion.

Projects

LangGraph ChatBot

LangGraph ChatBot

Multi-agent orchestration with LangGraph. Reduced tool-call errors −28%, task success +18%.

  • LangGraph
  • LangChain
  • Tool-Calling
  • Eval Suite
  • FastAPI
View on GitHub
Android agent

LLM-Based Android Agent

NL → UI actions with memory & self-reflection; hit 80%+ step accuracy on 10+ apps, goal success up ~25%.

  • Android
  • LLM Agent
  • Few-shot
  • Self-Reflection
  • Computer Vision
View on GitHub
Proxy TuNER

Proxy TuNER

Proxy-tuning for BERT (logit ensembling + GRL): avg F1 +8%, compute −70%, inference +30%.

  • BERT
  • Domain Adaptation
  • GRL
  • Efficient Tuning
  • PyTorch
View on GitHub
IntelliMeet

IntelliMeet

Decentralized video with edge ML (RetinaFace + Transformer STT): <200 ms latency, engagement +25%, 99.9% uptime.

  • WebRTC
  • Edge ML
  • RetinaFace
  • Transformer STT
  • E2E Encryption
View on GitHub
DNA classifier

DNA Sequence Classifier

DNABERT & HyenaDNA (LoRA + soft prompts); automated 1M+ seq preprocessing with Airflow/Biopython; accuracy 94%+, preprocessing time −40%.

  • DNABERT
  • HyenaDNA
  • LoRA
  • Airflow
  • Biopython
View on GitHub
PulseMap

PulseMap — Community & Disaster Intelligence

Fused USGS/NWS/EONET/FIRMS with AI-classified reports into one map. Polygon→point for 100% mappability; geo/time de-dupe; faster local awareness vs. juggling 4 sources.

  • LangGraph
  • Python
  • SQL
  • Docker
  • Geo Pipelines
  • Data Fusion
  • Map UI
  • Photo Classifier
View on GitHub

Background

My journey blends hands-on machine learning, data engineering, and product thinking.

  1. Jul 2025 – Present Experience

    ML Engineer — Cloud Systems LLC

    Built and shipped lightweight LLM agents for data retrieval and workflow automation, improved SQL/ETL performance by ~25%, and designed a hybrid retrieval stack (FAISS + BM25 + cross-encoder) that new client teams adopted within the first quarter, eliminating ~15 engineer-hours per week and accelerating pilot delivery.

  2. Sep 2024 – Jul 2025 Experience

    ML Research Engineer — Virginia Tech

    Delivered 94%+ DNA sequence classification using LoRA + soft prompting, automated preprocessing for 1M+ sequences with Biopython/Airflow to cut runtime 40%, and stood up semantic search over genomics literature—culminating in two publications (IEEE BIBM 2024; ML in Computational Biology 2025) and faster iteration cycles for the lab.

  3. Jan 2023 – Dec 2024 Education

    MEng, Computer Science — Virginia Tech

    M.S. in Computer Science (GPA 3.95/4); focused on AI/ML and data systems, co-authoring two papers on DNA foundation models and circadian transcription while building end-to-end LLM pipelines from training to deployment.

  4. Jun 2023 – May 2024 Experience

    Research Assistant — Virginia Tech

    Scaled genomic ETL with Airflow to boost data availability ~50%, automated retraining/eval loops to reduce manual work 40%, and optimized cluster workloads to trim runtime/resources ~15%, enabling experiments to grow from ~100K to 1M+ samples without additional hardware.

  5. Jul 2021 – Dec 2022 Experience

    Data Engineer — UJR Technologies

    Moved batch ETL to real-time streams on Kafka/Spark to cut processing latency 30%, containerized microservices on AWS ECS for 25% faster releases, and redesigned Snowflake schemas/materialized views for 40% quicker queries—sustaining 99.9% uptime across three enterprise environments.

  6. Aug 2018 – May 2022 Education

    B.Tech, Computer Science — Vel Tech University

    B.Tech in Computer Science (GPA 8.24/10); completed core systems and ML coursework and led a distributed-systems capstone that evolved into a small open-source contribution adopted by student developers.

Agents & LLMs

Snapshot • used weekly

Last used: this week

Data & MLOps

Pipelines • evaluation • observability

Last used: this week

Backend & APIs

FastAPI • auth • testing

Last used: this week

Cloud & Infrastructure

Docker • K8s • AWS/GCP

Last used: recent

ML / Analytics Tools

PyTorch • TF • CV/NLP

Last used: recent

Let’s talk

Open to roles and collaborations. The fastest way to reach me is email or LinkedIn.

dhulipallakrishnavamsi@gmail.com LinkedIn