GO
Machine Learning Engineer · Apple Intelligence

Machine Intelligence, engineered.

I'm Dhanush Balakrishna — building

View CV See the work
SCROLL
Selected impact

Under 10 seconds, on-device.

Agentic Hybrid RAG running entirely on-device — a Convex retrieval sweep resolves complex queries in under ten seconds.

2 weeks → 2 minutes.

Nexus — a multi-agent GenAI compliance platform across 37 regions and 190 countries. Natural-language rule discovery, MCP-backed and audit-ready.

Rooftops, read from space.

Hyperion — satellite-imagery computer vision for solar: +30% detection accuracy, +40% throughput, +50% estimation efficiency.

97% accuracy, 3× smaller.

Pruned & quantized a YOLOv5 detector from 240MB to 80MB and cut runtime 67% for real-time perception on edge devices.

100K+ frames, taught to see.

An automated pose-estimation pipeline over 100K+ infant images — 50% less manual labeling — powering invariant-representation research at Northeastern.

About

An ML engineer who ships from research notebook to production.

Dhanush Balakrishna

I'm a Machine Learning Engineer on the Apple Intelligence Proactive team at Apple, with an MS in Electrical & Computer Engineering from Northeastern University, specializing in Computer Vision and Natural Language Processing. My work centers on applied AI and agentic systems.

At SolBid I built the SolBrain API for solar cost estimation and led the Hyperion AI engine that detects rooftop obstacles from satellite imagery. Across PayPal, SolBid, British Airways and SandLogic, I've designed and deployed training and inference pipelines — rooftop & obstacle detection, segmentation, face recognition, and GANs — across cloud and edge platforms.

My toolkit spans PyTorch, TensorFlow, Keras, OpenCV, MCP, A2A, and I program in Python, C/C++, SQL, R, PHP Laravel, MATLAB and ROS. Outside the terminal I'm into photography, music production, and California coastal drives.

Location
San Jose, CA
Focus
CV · NLP · GenAI
Email
Get in touch →
Languages
EN · HI · KN
0
Years building ML systems
0
Faster rule discovery on Nexus
0
Posture images curated
0
Shipped projects
Specializations

What I work on, end to end.

Deep Learning

Architecting, training and optimizing neural nets across vision and language tasks.

Computer Vision

Detection, segmentation, pose estimation and depth — from satellite to edge devices.

NLP & GenAI

RAG, agentic orchestration, MCP/A2A, machine translation and LLM tooling.

Data Systems

ETL pipelines, SQL/NoSQL stores, and audit-ready data flows at scale.

Cloud & MLOps

AWS, Docker, GitHub Actions — reproducible deployment for training & inference.

Engineering

FastAPI, LangGraph, productionizing models behind clean, testable APIs.

Data Analytics

Sentiment, forecasting and dashboards that turn signals into decisions.

Web & Design

Building the interfaces that make models usable for real people.

Experience

A track record across research & industry.

Machine Learning Engineer

Present
Apple · Apple Intelligence (Proactive)
  • Building applied machine-learning and agentic systems for the Apple Intelligence Proactive team.
  • Work spans model evaluation, deployment tooling and large-scale ML systems.

Machine Learning Software Engineer

2025
PayPal Inc.
  • Led Nexus, a multi-agent GenAI compliance platform spanning 37 regions and 190 countries; enabled natural-language rule discovery with full CRUD, cutting discovery time by ~90% (from ~2 weeks to 2–3 minutes).
  • Architected dual access paths — a semantic RAG web app and a Claude-powered CLI (MCP client + custom servers) enabling Git-backed traversal and stateless subtask agents; unified 4 disparate sources.
  • Built scalable, audit-ready pipelines with LangGraph agents for classification, retrieval and validation. Stack: Python, FastAPI, LangGraph, Claude, FastMCP, Docker, GitHub Actions.

ML Research Engineer — Computer Vision

2024
Augmented Cognition Lab · Institute for Experiential AI, Northeastern
  • Engineered an automated annotation pipeline using MediaPipe's PoseLandmarker to extract infant posture keypoints, reducing manual labeling time by 50%+.
  • Curated 100K+ infant posture images from baby-monitor video, auto-annotating in bulk via a custom Python script.
  • Improved infant pose-estimation accuracy by refining ingestion workflows, integrating segmentation masks, and iterating on an invariant representation-learning model.

AI/ML Engineer

2024
SolBid Inc.
  • Boosted solar-project estimation efficiency 50% by productionizing ML algorithms (Python, Django) with a Glue ETL pipeline into MySQL on a Hadoop-Spark architecture.
  • Increased decision-making efficiency 40% by architecting financial analytics APIs in Laravel for ROI and payment scenarios.
  • Raised roof-detection accuracy 30% and processing speed 40% using ResNet, OpenCV imaging, optimal panel-placement algorithms, and AWS S3 data management.

Data Science Intern

2023
British Airways
  • Ran sentiment analysis on 15,000 customer reviews (BeautifulSoup, Pandas, NLTK, VADER), surfacing that ~20% of comments focused on in-flight amenities.
  • Built Random Forest & XGBoost models for rebooking prediction, lifting accuracy from ~60% to ~78%.
  • Designed interactive dashboards (Seaborn, Power BI, Tableau), cutting manual review time ~30%.

Machine Learning Research Engineer

2021
SandLogic Technologies
  • Supervised a team of six building a road-obstruction detection model (CNN MobileNet) at 80% accuracy on dashcam footage for autopilot applications.
  • Improved annotation efficiency 50% via AuVi.io, Edgematrix.io and Makesense.ai tooling.
  • Cut runtime 67% and hit 97% accuracy on edge devices with TensorFlow Lite; reduced model size 40% (240MB → 80MB) via YOLOv5 pruning & quantization.
Education

MS, Electrical & Computer Engineering

2022–2024
Northeastern University · ML, CV & Algorithms
  • Coursework: Algorithms, Computer Vision, Advanced Perception, Machine Learning, Robotic Science & Systems, Foundations of AI, Advances in Deep Learning, DBMS.
  • Teaching Assistant — ML & Pattern Recognition, Applied Probability & Stochastic Processes, Information Theory.

BE, Electronics & Communication

2018–2022
Visvesvaraya Technological University
  • Coursework: Image Processing, Web Programming, Data Structures, OOP, Computational Mathematics.
Skills

The stack I reach for, daily.

Languages

Python95%
C / C++82%
SQL86%
R / MATLAB74%

ML & GenAI

PyTorch93%
TensorFlow / Keras86%
OpenCV88%
LangGraph / MCP / RAG85%

Backend & Infra

FastAPI88%
Docker85%
AWS82%
MySQL / PostgreSQL84%
Selected work

Projects across vision, language & systems.

Python TensorFlow PyTorch AWS MySQL R jQuery Laravel PostgreSQL OpenCV

Let's build something intelligent.

Always happy to talk shop — research collaborations, ML systems, and interesting problems.

Location
San Jose, CA, USA
Phone
+1 (617) 817-6535
Schedule
Book a call →