AI agents & automation
Agents and LLM workflows that plan, use your tools, and get evaluated before they reach users.
multi-agent systems · knowledge graphs · agent memory · agent evaluation
AI consulting and implementation
Senior AI delivery you can put in front of a client.
Practical AI, taken from prototype to production and handed over running. You work directly with the engineer who builds it: no junior handoff, no account-manager buffer.
Track record you can check
A senior practitioner, not a pitch.
Ten-plus years building and shipping production AI, from PhD research to a funded startup as Chief AI Officer. Here is the arc, and the evidence.
- 2026FounderNazmi
- 2025-26Principal AI ScientistIn Parallel
- 2022-25Co-founder & Chief AI OfficerResoniks
- 2021AI engineerSilo AI (acquired by AMD)
- 2019-nowPhD research, ML & roboticsAalto University
- 10+years building production AI & ML
- 100+technical articles written
- €2.65Mseed round co-founded
- Recommender systems and AI agents for a retail-technology engagement.
- ML workflows on Azure Machine Learning and Databricks for an enterprise telecom engagement.
- RAG, MLOps, and anomaly detection across support, healthcare, and energy.
You work directly with Isaac as principal on every engagement. When scope needs it, Nazmi scales with a small network of vetted senior collaborators — never a junior handoff.
See selected workWhat Nazmi does
Three fields Nazmi takes to production.
Not a broad menu — three areas where Nazmi ships, each backed by systems in production, published writing, or research you can read.
Production ML & MLOps
Models out of the notebook and kept alive in production — monitored, repeatable, owned by your team.
Azure ML · AWS · GCP · Databricks · MLOps · recommenders · ranking · anomaly detection
Private & on-prem AI
Capable AI inside your own walls — small, private models for when data cannot leave and API bills will not scale.
private LLMs · on-prem deployment · fine-tuning · RL optimization · efficient inference
How Nazmi works with you
Bring Nazmi in the way that fits.
Most work comes in through partners and recruiters who need AI hands they can trust with a client — plus direct company projects.
For delivery & technology partners
AI implementation capacity for your client projects, from architecture to deployment, without adding headcount. Brought in as a peer.
For recruiters & staffing
A senior AI practitioner for contract, fractional, and advisory placements.
For companies
Practical AI projects with a real workflow, owner, and data path, taken from idea to a system your team can run.
Selected writing
Depth you can read.
A selection from 100+ technical articles on agents, RAG, small-model post-training, and inference. The thinking behind how Nazmi ships.
AI Agents in Production
Before an AI Agent Touches Customer Data: Six Production Gates It Must Pass
A demo proves an agent can sometimes do the job. Production requires evidence it can operate safely, predictably, and recoverably. Six gates every team should pass before letting an agent near real data.
Read →Context Compaction in LLM Agents
Context Compaction in LLM Agents: Part 2 - Learning to Compact
The next frontier of context compaction is not a better heuristic but a model that decides for itself when, what, and how to compress. A look at SelfCompact, CompactionRL, and ACON.
Read →Context Compaction in LLM Agents
Context Compaction in LLM Agents: Part 3 - Post-Hoc Compilation
Every compaction method we have covered so far works within a single run. Lerim takes a different approach: wait until the run is done, then compile durable signal for the next agent.
Read →Context Compaction in LLM Agents
Context Compaction in LLM Agents: Part 4 - Theory and Safety
Is there a unified way to think about all compaction methods, from KV cache eviction to agent memory? And what happens when compaction silently erases safety constraints?
Read →Small-Model Distillation
Small-Model Distillation — Part 5: On-Policy Self-Distillation for a 0.8B SQL Agent
I had a 0.8B SQL agent scoring 69/220 on the fixed eval split, built through four stages of distillation from larger teacher models. The question was whether…
Read →Small-Model Distillation
Small-Model Distillation — Part 4: On-Policy Probability Distillation for a 0.8B SQL Agent
I started this post from the best hard-token correction-family SQL agent I had so far: unsloth/Qwen3.5-0.8B, evaluated at 67/220 on the fixed eval set. That…
Read →Work with Nazmi
Bring Nazmi in on the work that has to ship.
Tell us what you are building — a project, a partner need, or a client problem. A few sentences is enough.
Start a conversation or book a 20-minute call →