Solution Architect — Aivar (AWS)
End-to-end work from pre-sales meets to designing and delivering agentic solutions to large enterprise ($1B+) clients.
Ram Kadiyala
Enterprise agentic AI, Gen AI research, and competitive esports.
Agents
Multilingual LLMs
LLM Post-training
Benchmarks & Evals
Background summary.
I was an early beta tester before agents were publicly released. Since then I've been the founding solution architect for agentic AI at two startups — one in the US, one in India. Over the weekends, I lead the ML-Agents group at Cohere Labs, and consult on agentic AI and multilingual post-training data for a few labs and startups. Prior to that, I spent three and a half years bootstrapping an esports startup.
Weekdays, weekends, and the in-between.
Open-source Lead — ZeroGrad.ai
Leading a team of open-source researchers at a non-profit applied research group based out of India.
Community Projects — Cohere Labs
Leading community projects utilizing a large pool of resources ($50K+) in collaboration with other organizations.
Get to know my research.
Compliance, observability, and governance for agentic workflows in production and customized agent stacks for high-value customers.
On the research side, I work on Auto-optimization of agents (AutoML for agents)
Speech and text models for Indian languages—OCR, TTS, STT, language detection, and region-of-origin detection. Work runs from data prep and training through to models that have to handle scripts, dialect mix, and real deployment constraints.
Optimizing results with fixed datasets in model training—post-training and alignment. LLMs for Indian-language coverage, On-device LLMs via distillation and quantization for edge and mobile inference.
Domain-specific agentic benchmarks and better reliable evals. Multilingual benchmarks with native, code-mixed, and transliterated versions.
Streaming tokens for lower latency while maintaining quality. Additionally, exploring world models, text-to-3D, and RL with browser agents.
Certifications.
Hover any card to flip it.
AWS ML Specialty
Valid till Mar 2029
AWS ML Specialty
Bedrock · SageMaker · AgentCore · Feature Store · Model Registry
CUDA Computing
12+
Papers contributed
9×
1st-place shared tasks
$50K+
Community compute led
200+
People led in esports
Education.
Hover any card to flip it.
M.S., Machine Learning
2022 – 2024
University of Maryland
M.S., Machine Learning
Certificate, Data Science
2022 – 2023
University of Maryland
Certificate, Data Science
B.Tech., Computer Science
2018 – 2022
RVR & JC College of Engineering
B.Tech., Computer Science
Experience.
Tap the + on any card to open the full story.
Aivar
Aivar · Nov 2025 – Present
AI Solution Architect
Leading solution design, building end-to-end production systems, and delivery of agentic and LLM-based systems for high-value enterprise clients on AWS across India, US and Middle East.
- PineLabs – Semi-autonomous SRE Agent: Self-healing incident resolution across L1–L3 tiers with HITL feedback loops and engineer annotation-driven continual improvement.
- HighRadius – Infra Resource Allocation Agent: Autonomous provisioning and right-sizing of cloud resources with demand-driven scaling and engineer-in-the-loop override controls.
- AGS Health – Healthcare Claims Appeal Agent: End-to-end automation of claims appeal workflows processing up to 30 document types and 300+ page inputs/outputs via multi-agent pipelines.
- Swiggy – Membership Conversion Forecasting: SageMaker-based pipeline with automated A/B testing and best-model selection to predict user membership purchase likelihood.
- M2P Fintech – Banking File Parser: One agentic pipeline to parse and standardize payment and banking documents across thousands of formats from multiple financial institutions across the globe.
Cohere Labs
Cohere Labs · May 2025 – Present
Open Science Community Lead
Lead the open-source research program across two groups – Community Projects and ML-Agents – covering LLM post-training and agentic systems at production scale.
- Program Leadership: Leading the open-science program across both groups, finding research gaps, setting direction, and coordinating distributed contributors. labscommunity.cohere.com/community-leads
- Grants & Funding: Secured over 50K+ USD in compute and funding grants from Anyscale, Hugging Face, and Modal sustaining community project requirements at production scale.
- Pre-Launch & GTM: Beta-testing feedback and go-to-market input on new Cohere model releases, bridging the labs community and product teams.
- Mentorship: Mentoring undergraduates and early-career researchers on their research ideas, with 12+ papers published through the program at tier-1 venues: ACL, EMNLP, NeurIPS, etc.
Traversaal.ai
Traversaal.ai · Sep 2023 – Aug 2025
NLP Research Lead
Delivered production agentic systems and AI products for Fortune 500 clients while leading NLP research on multilingual models and agent evaluation benchmarks.
- Home Depot – Route Optimization Agent: Built an agent for delivery route planning that incorporates user-specified constraints such as time windows, evening-only deliveries, and blackout periods.
- Fortune 500 Clients – Agentic OS: Developed an organization-wide agentic operating system for data access, workflow automation, and orchestration; deployed in production for multiple Fortune 500 companies.
- Enterprise Clients – Web Scraper & Search Engine (Ares): Built a scalable web scraping and semantic search engine for enterprise data ingestion and retrieval; adopted by several large organizations.
- Research Initiatives – Multilingual LLMs & Benchmarks: Curated synthetic datasets for Indian languages with localized context; developed benchmarks and frameworks for fine-grained evaluation and continual learning of domain-specific agents.
- Course Contributor: Contributed to curriculum development for the “Building and Scaling AI Solutions” course, Spring 2025, Stanford Continuing Studies.
Black Ops Esports
Black Ops Esports · Nov 2018 – Apr 2022
Founder
Founded and scaled a mobile-gaming esports organization spanning competitive teams, a 60K+ community, a monetized regional creator network, and data-driven predictive modeling for in-game strategy.
- Community Growth: Grew the organization’s social presence organically with $0 ad spend to 56K on Instagram and an 18K member Discord community.
- Creator Network & Monetization: Built and managed a network of regional gaming content creators across multiple Indian languages – connecting them with advertisers, guiding content strategy and collaborations, and earning a share of the ad revenue generated.
- Competitive Teams: Recruited and managed competitive esports teams of 200+ members across several mobile gaming titles.
- Game Analytics: Built game-progression forecasting models for battle-royale titles to simulate match dynamics and stress-test drop and rotation strategies against projected placement outcomes.
In good company.
Open as many as you like — they stack.
NLP Research Lead (Sep 2023 – Aug 2025). Led a team of 15 on vertical agent applications, multi-agent frameworks, multilingual LLMs for South-Asian languages, and fine-grained agent evaluation.
Community Projects Lead (May 2025 – Present). Production-scale ($50K+) research on LLM post-training and agentic applications with partners like Anyscale and Modal.
AI/ML Solution Architect (Nov 2025 – Present). Solution design and delivery of agentic and LLM-based systems for large AWS clients.
Lead a team of open-source researchers at a non-profit applied research group based out of India.