Hello, I'm Suhas
Transforming complex challenges into elegant solutions across AI, software engineering, and cloud infrastructure
Engineer. Researcher. Mentor.
- Proven expertise in AI research, cloud-native software, and teaching excellence.
- Design intelligent, scalable software that blends rigorous problem-solving with real-world application.
- Navigate across abstraction layers—from high-level design to low-level optimization.
- Analytical thinker: reduce complexity, build reliable solutions, and value clarity, efficiency, and continuous learning.
Highlights
- 3+ years shipping backend/data systems
- Cloud Certified (AWS SAA + CP)
- Led LLM alignment research at Northeastern
- Hackathon winner | AiXplain 2024
- 40+ students mentored as Teaching Assistant
- Full-stack development experience
Experience
Graduate Research Assistant
Northeastern University
- Led research on the Adaptive Incremental Learning Architecture (AILA)
- Designed and implemented a novel attention mechanism, improving sequence modeling accuracy by 12%
- Achieved 15% faster training times while maintaining model performance
- Contributed to Large Language Model (LLM) post-training with RLHF
- Created enhanced reward models that improved alignment with human preferences
Graduate Teaching Assistant
Northeastern University
- Mentored 40+ graduate students in advanced Data Science topics
- Provided hands-on guidance for MLOps implementation and deployment pipelines
- Conducted weekly lab sessions focused on practical ML applications
- Held regular office hours to support student learning and project development
- Created supplementary learning materials that improved project quality by 30%
- Developed comprehensive assignment rubrics for consistent evaluation
Cloud Engineer Intern
Safecast
- Optimized AWS resource allocation across multiple service tiers
- Implemented auto-scaling strategies reducing monthly cloud costs by 35%
- Maintained all performance metrics while decreasing infrastructure expenses
- Migrated legacy services to a modern microservices architecture
- Utilized containerization and orchestration for improved system reliability (25%)
- Established comprehensive infrastructure documentation
- Created disaster recovery protocols that decreased incident response time by 40%
Software Engineer
Runway Proptech LLC
- Led backend development for a property management platform serving 10,000+ users
- Architected RESTful APIs and microservices using Python and AWS
- Developed scalable database solutions with robust data validation
- Integrated AI-powered recommendation services increasing client retention by 22%
- Built custom forecasting tools for property performance analysis
- Implemented CI/CD pipelines reducing deployment time by 60%
- Created comprehensive monitoring solutions that decreased critical incidents by 35%
Projects
AlignAI
Developed an AI alignment platform that improves LLM outputs through RLHF and advanced reward modeling, increasing response quality by 32%.
- PyTorch
- RLHF
- Reward Modeling
- Transformers
- Fine-tuning
QueryMaster AI
AI app to achieve over 90% accuracy in translating natural language into SQL queries, enhancing data accessibility for non-technical users.
- Python
- LangChain
- OpenAI
- Vector DB
- FastAPI
SafeHomeSeeker AI
DC SafeHouseFinder AI: Built during aiXplain's Hackathon, provides personalized housing recommendations using AI and geospatial tech.
- AI Agents
- Vector DB
- aiXplain
MatchPoint: The Resume Analyst
Developed a resume scoring app using React and FastAPI, Dockerized deployment on AWS ECS, providing real-time NLP-based feedback and suggestions
- React
- FastAPI
- AWS ECS
- Llama Model
- Docker
Nutrition Tracking Database
Developed a Nutrition Tracking System with SQL Server, Tableau dashboards, and secure handling of user-specific dietary and fitness data.
- T-SQL
- Database Design
- Tableau
- Data Security
Research
AILA: Adaptive Integrated Layered Attention
Authors: William Claster, Suhas K M, Dhairya Gundechia
Proposed a novel neural architecture for adaptive integrated layering of attention mechanisms, enabling efficient and effective integration of multiple modalities in AI models.
REALM: Enhancing Reward Models for LLM Alignment introducing Explicit Prompt-Answer Relevance
Authors: William Claster, Suhas K M, Dhairya Gundechia
An alignment learning model for relevance-enhanced alignment of AI models, enabling improved performance and accuracy in AI applications.
Skills
programming
cloud
databases
Data Science & AI
tools
interpersonal
Contact me
Fill out the form below and I'll get back to you as soon as possible.
Location
Washington D.C
Open to relocate for on-site opportunities