We build software and AI products end to end, from prototype to production.
Techlon is a product engineering team for startups and small companies. We take an idea, a rough spec, or a stalled project and turn it into something real, shipped, and running in production — the design, the code, the models, and the infrastructure underneath.
01 — Capabilities
Four things we do. Most projects use more than one.
01
Product Development
We start before the code. Scoping, user flows, architecture decisions, and a build plan that survives contact with reality. If you know the problem but not the shape of the solution yet, this is where we begin.
02
Web & App Development
Production web applications and mobile apps. Modern stack, tested, documented, and handed over in a state your own engineers can pick up. No black boxes.
03
Data Science
Getting answers out of data you already have. Pipelines, analysis, forecasting, dashboards your team will actually open. We start with the decision you're trying to make and work backwards.
04
AI & Model Building
Custom models, LLM applications, RAG systems, and evaluation to prove they work. We build AI that ships to real users — not demos that fall apart the first week under load.
02 — How we work
Custom software goes wrong in predictable ways. Here's how we avoid them.
We scope before we quote.
Fixed prices on vague requirements are how projects get 40% built and abandoned. We do paid discovery first, then quote against something real.
You see working software every week.
Not status reports. A deployed environment you can click through, from week one. If we're going the wrong direction, you'll know in seven days instead of three months.
One team, start to finish.
The people on your first call are the people writing the code. No handoff to a junior team after you sign.
The bill is never a surprise.
Fixed-price sprints and a running total you can check any time. If something threatens the budget, you hear it from us first — before it costs you anything.
03 — Discovery sprint
Not sure what you need built yet? Start here.
Two weeks, fixed price. We dig into the problem, map the technical approach, and hand you a complete build plan: architecture, scope, timeline, and a fixed quote for the full project.
You keep the plan whether or not you build with us. If you do continue, the sprint fee comes off the project cost.
A discovery sprint is normally $2,000. For the next three months we're running it at 75% off, while we build our first public case studies. It's the cheapest way to find out whether this is a three-month project or a nine-month one, before you've committed to either.
$2,000 75% off · 3 months
04 — Work
Things we've made
2026 · Adj. R² 0.9250
Sales Cannibalization
A nested logit model to find the primary drivers of a company's sales — and what share of a new product's sales came from cannibalizing existing ones.
2026 · ROC-AUC 0.761 · Acc 0.744
Fake Review Detection
An ensemble of SBERT, XGBoost, and RoBERTa that detects fake product and service reviews, with an LLM auditor checking the model's calls.
2025 · No significant lift
A/B Testing
A one-tailed t-test on whether a new page increased conversion. It didn't — and the null held across every heterogeneity analysis. Knowing that before rollout is the point.
2023 · Accuracy 0.9875
Image Recognition
A convolutional neural network for object classification, built with TensorFlow and Keras.
2022 · F1 0.847
Credit Card Fraud Detection
A fraud pipeline in Scikit-learn comparing XGBoost, Random Forest, SVM, decision trees, and logistic regression on heavily imbalanced transaction data. Scored on F1, not accuracy — on data this skewed, accuracy flatters a model that never catches anything.
05 — Team
The people who'll be building it
Techlon is two engineers based in Cleveland, Ohio. We started Techlon because [real reason, one sentence].
Oluwagbemi Lesi
Founder
MBA in Business Analytics from Case Western Reserve, on top of seven years across financial services, real estate, and FinTech regulation. Builds and validates predictive models in Python — forecasting, credit risk, NLP — and specializes in model validation: finding where reported performance overstates real signal. Before Techlon: head of operations at a FinTech checkout platform, co-founder of a real estate firm, and a lawyer designing compliance frameworks for regulated financial products.
Emeka Osuagwu
[Role]
[Profile and photo coming.]
Want to be one of them? See how we work and what we value →
06 — Questions
We're a startup — is this going to be too expensive?
Depends what you need. An MVP scoped to prove one thing is a different project from a full platform, and we'll tell you which one you're actually asking for. The discovery sprint gives you a real number before you commit to anything.
We have an internal team already.
Plenty of our clients do. We come in for the piece the team doesn't have capacity or specialist depth for — an AI feature, a data pipeline, a rebuild of something creaking — and we work in your repos, your standards, your review process.
Techlon is new. Why should we take the risk?
Fair question. Two answers: our work is public, so you can inspect it before you talk to us. And we scope in two-week paid increments, so you're never exposed for more than one sprint — if it isn't working, you stop, and you keep the build plan.
How fast can you start?
Usually within two weeks. Tell us your timeline on the call and we'll be straight with you about whether we can hit it.
Tell us what you're trying to build.
Thirty minutes, no pitch deck. Bring a problem, a half-formed idea, or a project that's stuck — you'll leave with a straight answer about whether it's worth building and roughly what it takes.