$ls ./work
Case studies, not screenshots
Four builds documented the way a technical interview goes, then the prototypes underneath them.
retrieval eval · run 214
passingnDCG@10
0.81
+.09
MRR
0.74
+.06
Recall@20
0.92
+.04
Berribot Retrieval Evaluation
Turning "the answers feel better" into a number the team could trust.
Applied AI Engineer, owned retrieval quality end to end
- nDCG@10 lift
- {{NDCG_LIFT}}
- Recall@20
- {{RECALL_AT_20}}
PHYS 216 · unit 4
7 / 12
Which quantity is conserved when the net external torque is zero?
source · lecture 11, slide 6
verified against source
ThirdSlate
A study platform that reads the actual course, not the internet.
Co-founder and Product Engineer, owned product direction and the retrieval stack
- Students served
- 1,000+
- Questions generated
- {{QUESTIONS_GENERATED}}
checkout
128.40 USDC
order #A-4417 · base
reconciliation
export → csv · matches order ledger
Outpay
Non-custodial USDC checkout that merchants can actually reconcile.
Solo build: product definition, contracts integration, frontend and merchant dashboard
- Settlement time
- {{SETTLEMENT_TIME}}
- Fee vs card rails
- {{FEE_DELTA}}
trace · task 88f2
4.2s · 6 callssuccess
0.86
tool prec.
0.91
cost/query
$0.03
Tekkscope
Agentic search infrastructure that fails loudly instead of quietly.
Solo build: architecture, agent orchestration, reliability harness
- Task success rate
- {{TASK_SUCCESS}}
- Tool call precision
- {{TOOL_PRECISION}}
02$ls ./playground -a
Things I built to find out
Prototypes and research. Not everything here shipped, and the state on each one is honest.
$./next-step
Hiring for a product engineer who can carry the whole thing?
Send me the role and the problem behind it. I will tell you honestly whether I am the right fit.